DEV Community: Daniel Marin

My Calendar Was Running My Life. These 4 AI Skills Gave Me My Week Back.

Daniel Marin — Sun, 28 Jun 2026 05:11:14 +0000

Protect focus blocks automatically, find availability in 30 seconds, surface hidden meeting dysfunction, and start every day with a plan instead of a scramble.

A packed calendar and a productive week are not the same thing. Most professionals who feel overwhelmed by their schedule are not overwhelmed by work. They are overwhelmed by the overhead that surrounds work: the scheduling ping-pong emails that consume 10 minutes per meeting request, the focus blocks that get overwritten the moment someone needs "just 15 minutes," the morning tab-opening ritual that eats the first 20 minutes of every day before a single real task has been touched, and the meetings that feel purposeless because the actual disagreements happened in Slack after everyone said "sounds good" in the room.

The calendar is the structure that the week runs on, but most people interact with it reactively: accepting what others put on it, manually managing conflicts, and starting each day without a clear picture of what needs preparation. The result is a week that happens to you rather than a week you designed.

The Calendar Overhead Problem

There are two distinct ways a calendar causes lost time. The first is direct: meetings that should not exist, meeting durations that are padded beyond what the agenda requires, back-to-back scheduling that eliminates transition time and makes every meeting start behind. These are the obvious problems.

The second is indirect and less visible: the administrative overhead of managing the calendar itself. Finding a time that works across three participants' schedules (cross-referencing calendars, accounting for time zones, identifying slots that do not collide with known focus blocks) takes 10 minutes for a meeting that will last 30. Multiply that by five meeting requests per week and the scheduling administration consumes an hour before the meetings themselves have started.

The third layer is meeting quality. An hour in a meeting that ends with fake consensus (everyone nodding while disagreements go unspoken, decisions made that get quietly reversed in side conversations) is not a productive hour. It is a productive-looking hour that produces no real alignment.

1. Calendar Automation: Protect Focus Blocks and Automate Meeting Workflows

This skill builds automated workflows for Google Calendar and Outlook: auto-blocking focus time that resists being overwritten, sending meeting prep documents before scheduled calls, posting daily agendas to Slack, and syncing across calendar platforms. So the administrative work of managing a calendar runs without manual intervention.

The focus block problem has a specific shape: most professionals who try to protect deep work time do so by manually blocking calendar slots, only to find that other people schedule over them. The person asking for "just 15 minutes" does not see a focus block. They see available time. This skill changes how focus blocks are created and protected: blocks that are marked as high-priority, that automatically decline meeting requests during protected hours, and that are automatically rebuilt if deleted. The calendar defense is systematic, not a daily manual battle.

"Set up calendar automation for my Google Calendar. I need: (1) daily focus blocks, 9am to 12pm Monday through Thursday protected as deep work, auto-decline any meeting requests during those hours with a polite message offering afternoon slots instead, (2) meeting prep automation, 30 minutes before any meeting over 45 minutes send me a prep note with the meeting agenda, attendees, and links to any relevant docs shared in the invite, (3) daily agenda post to Slack at 8am, (4) auto-sync between my Google Calendar and Outlook so both stay current."

Before: Monday 9am focus block: overwritten by Tuesday. New block created: overwritten Thursday. Every deep work session is a negotiation. Meeting prep: 8 minutes before each call opening email threads and project docs.

After: Morning focus blocks defended automatically. Requests during those hours receive a polite redirect to afternoon. 30-minute prep notes arrive before every substantive meeting. 8am Slack post: "Today: 10am Product sync (agenda: Q3 roadmap, prep doc attached), 2pm Client call (Smith account, last touchpoints: [links])." Zero manual overhead.

Setup: 10 minutes. The focus block auto-decline is the feature that converts calendar protection from a daily manual task into a system that runs without attention.

2. Availability Checker: Find the Right Slot in 30 Seconds

This skill checks Google Calendar across all calendars simultaneously, finds optimal meeting slots that respect focus blocks and buffer times, and drafts a professional scheduling reply with three to five time options. Converting a 10-minute per-request task into a 30-second one.

The scheduling ping-pong problem: "when are you free next week?" triggers a manual cross-referencing process. Opening the calendar, scanning for open slots, mentally checking whether those slots are actually good (not immediately after another intense meeting, not during the window reserved for deep work, not too late in the day to get value from a 60-minute call). Then writing a reply that lists the options in a format that does not create confusion about time zones. That process, repeated for every meeting request, accumulates into a significant hidden time cost.

"Check my availability for a 60-minute meeting next week. Rules: no meetings before 9am or after 5pm, protect my Monday and Tuesday morning focus blocks (9am to 12pm), require a 15-minute buffer between meetings, prefer not to schedule on Friday afternoons. Find 4 to 5 optimal slots and draft a professional reply to send to the organizer. Include time zone (Eastern) and a note that I'll send a calendar invite once they confirm."

Before: "When are you free next week?" Open Google Calendar, open Outlook, cross-reference, find a slot, check if there's buffer on either side, check the time zone difference, write the reply. 12 minutes. Do this five times a week: one hour per week on scheduling emails alone.

After: Four optimal slots identified in 20 seconds. Professional reply drafted, buffer verified, focus blocks respected, Friday afternoon avoided. Review, paste, send. 90 seconds total. No conflicts. No follow-up to clarify time zones.

Setup: 10 minutes (requires Google Calendar integration). Buffer time rules and focus block protection are configured once and applied automatically to every availability check.

3. Meeting Conflict Analyzer: Surface What the Meeting Did Not Say

This skill processes meeting transcripts to identify communication patterns that cause meetings to fail silently: conflict avoidance, unspoken concerns inferred from hedging language, communication dominance patterns, and decisions that appear made but are not actually aligned on. It produces coaching recommendations for healthier team disagreement.

The silent meeting failure problem: the meeting ends, everyone said the right things, and then three people message separately to say they disagree with what was decided. Or the decision gets made in the meeting and reversed in the back-channel conversation that happens afterward. The agenda was followed. The time was spent. Nothing was accomplished because the actual disagreements never surfaced.

Analyzing a meeting transcript for communication patterns reveals what the recording does not make obvious in real-time: the hedge ("that could work...") that signals unspoken reservation, the topic that got abbreviated when a senior person spoke, the concern raised and then immediately withdrawn when it was not validated.

"Analyze this meeting transcript for communication patterns. [paste transcript] I'm looking for: (1) instances of conflict avoidance, where someone raised a concern and then backed off, (2) unspoken concerns, hedging language or topic changes that suggest something was left unsaid, (3) dominance patterns, who spoke most, whose suggestions were built on versus ignored, (4) decision quality, which decisions appear made but may not have genuine buy-in. Give me coaching recommendations for how to run this meeting differently next time."

Before: Thursday team meeting: everyone nods, "sounds good," meeting ends. Friday: three separate Slack messages from attendees explaining why the decision will not work. Monday: the decision is informally reversed. Total time lost: 1-hour meeting + 2 hours of back-channel conversation + 1-hour re-decision meeting. The Thursday meeting accomplished nothing.

After: Transcript analysis: 3 instances of conflict avoidance identified, including one engineer who raised a technical concern and backed off when the PM pushed back. Dominance pattern: one person spoke 60% of the time, two attendees said nothing after the first 10 minutes. Coaching recommendation: start with silent written input before discussion. Next meeting: actual alignment reached. No back-channel reversals.

Setup: 10 minutes. Works with any meeting transcript. Most useful run quarterly on recurring meetings to track whether communication patterns are improving over time.

4. Daily Morning Briefing: Start the Day Oriented, Not Scrambling

This skill generates a unified daily view in 30 seconds: today's meetings with prep context, VIP emails needing same-day responses, priority tasks, and the top focus for the day. Consolidating what currently requires 20 minutes of tab-opening into a single brief read at the start of the morning.

The morning orientation problem: most professionals start their day by opening the same set of tools in sequence (calendar, email, task manager, Slack) and mentally assembling a picture of what the day requires. That assembly takes 15 to 20 minutes and happens in a fragmented, reactive way. The synthesis ("this is what actually needs my attention today, in this order, with this preparation") never gets done explicitly. The day is reacted to rather than planned.

This skill does the synthesis automatically. Pulling from Google Calendar, Gmail, and connected task tools, it produces a single view: meetings for the day with prep notes, VIP emails flagged for same-day response, tasks due today sorted by priority, and a suggested top focus for the day based on deadline pressure and meeting load. The 20-minute morning ritual becomes a 2-minute read.

"Generate my morning briefing. Pull from: Google Calendar (today's meetings), Gmail (emails from VIP senders or marked urgent that arrived since 6pm yesterday), task list (items due today or overdue). Format: (1) Today's schedule with time, duration, attendees, and one-sentence prep note, (2) Inbox priorities with sender and subject, (3) Task priorities sorted by urgency, (4) Top focus: the single most important thing to accomplish before my first meeting. Keep it to what fits on one screen."

Before: 7:45am: open calendar, scan meetings. Open email, scan for urgent flags. Open Slack, scan for overnight messages. Open task manager, figure out what is due. 8:05am: still not sure what needs to happen first. First meeting is at 9am and the prep notes are still not pulled. Day starts reactive.

After: 7:45am briefing received: 4 meetings today, prep notes for each. 2 VIP emails need responses. 3 tasks due, ranked by priority. Top focus: finish the Q3 analysis before the 10am stakeholder review. 7:47am: working on the Q3 analysis.

Setup: 15 minutes (requires Gmail and Google Calendar integration). The VIP sender list is configured once.

The Calendar System: Protect, Find, Improve, Brief

The four skills address calendar management at four levels. Calendar Automation handles the structural layer: protecting focus blocks, automating prep workflows, and keeping calendars synchronized. The Availability Checker handles the scheduling overhead layer: eliminating the 10-minute per-request cost of finding times. The Meeting Conflict Analyzer handles the meeting quality layer: making the hidden communication patterns visible so the hours spent in meetings actually produce alignment. The Morning Briefing handles the orientation layer: synthesizing the day's demands into a single view before the reactive tab-opening ritual can consume the first 20 minutes.

What the skills do not replace is the judgment about which meetings should exist at all. Calendar automation can protect focus blocks, but it cannot decide whether a recurring meeting has outlived its purpose. Availability checking can find the optimal slot, but it cannot decide whether the meeting is worth scheduling. That judgment remains with the person managing the calendar. What the skills handle is everything else.

Getting Started

I publish all four calendar skills as free, downloadable templates at claudecodehq.com: focus block protection with meeting prep automation, instant cross-calendar availability checking, meeting transcript analysis for communication patterns, and automated daily morning briefings. Each one is a single file you install once and use whenever scheduling chaos returns. Start with the Calendar Automation skill if your focus blocks keep getting overwritten, or the Morning Briefing if your days start with 20 minutes of reactive tab-opening.

Originally published on claudecodehq.com

I Had 2,347 Unread Emails. I Hit Inbox Zero in 90 Minutes. Here's the 4-Skill System.

Daniel Marin — Fri, 26 Jun 2026 09:29:37 +0000

Clear the backlog, automate ongoing triage, classify any batch by urgency, and draft the hard emails in 30 seconds instead of 20 minutes.

The unread count is the first thing most people see when they open their phone in the morning. For a lot of people, it has stopped being a useful number. It is just a number that is always larger than yesterday. 2,347. Then 2,401. Every time a real attempt is made to clear it, fifty emails get processed before the effort collapses under its own weight.

The problem is not laziness. The problem is that a modern inbox is an undifferentiated pile: client urgencies and newsletter editions and automated receipts and LinkedIn notifications and a thread from three months ago that technically needs a reply all exist at the same priority level until a human sorts them.

The inbox-as-to-do-list is the trap. When email is used as a task management system (which it is, for most professionals who have not explicitly built another system), the inbox has to be kept visible at all times so nothing important gets missed. That means checking it constantly. That means every low-priority notification interrupts focused work. That means the reply to the sensitive client escalation gets written at the same distracted speed as the unsubscribe to a newsletter.

Four AI skills address the email problem at all four pressure points: clearing an existing backlog without burning a full day on it, building an ongoing triage pipeline that sorts incoming mail automatically, classifying what actually needs human attention versus what can be archived, and drafting the replies that take too long to write from scratch.

Why Email Overwhelm Compounds

Email volume has a compounding problem. Every day that the backlog is not processed, it grows. Every day it grows, the prospect of clearing it becomes more daunting. Every day it becomes more daunting, the probability of attempting it decreases. By the time the inbox hits 2,000 unread, most people have made a tacit decision to abandon it, checking only for obviously urgent items, letting the rest accumulate, and living with the background anxiety of the number.

The second compounding problem is sorting cost. In an unsorted inbox, reading email requires reading every email. There is no way to know whether the next item is a client escalation or a promotional newsletter until you open it. That friction (the cognitive cost of constant re-prioritization) is what makes processing 500 emails feel like it takes five times as long as processing 100. The work is not reading. The work is sorting while reading.

The third problem is drafting friction. Some emails require writing. A reply to a difficult client, a pushback on scope creep, a sensitive performance note, a follow-up after a missed deadline. These emails take 20 minutes not because they contain 20 minutes of information but because the tone must be calibrated carefully and the blank page provides no starting point.

1. Inbox Zero Manager: Clear the Backlog Without Burning a Day

This skill processes an email backlog of any size (500, 2,000, or more) categorizing every email into urgent, action-required, FYI, and archive buckets, drafting responses in your voice for everything that needs a reply, and flagging bulk-archive candidates so newsletters and automated notifications can be cleared in seconds. Nothing gets sent without your explicit approval.

The backlog clearance problem has a specific shape: most people know they need to process old emails, but the activation energy required to start is high because there is no visible end point. Starting means committing to an unknown amount of work. This skill changes that by front-loading the categorization. Before any reply is written, every email gets sorted. Once the sorting is done, the pile becomes visible: 15 genuinely urgent items, 30 that need a reply, 200 newsletters to bulk-archive, 300 automated notifications to delete. The work is now bounded.

"I have 600 unread emails I need to process. I'm going to paste them in batches. For each batch: categorize each email as URGENT (needs same-day response), ACTION (needs a reply within a week), FYI (read but no reply needed), or ARCHIVE (newsletters, receipts, notifications, automated mail). For the URGENT and ACTION emails: draft a reply in my voice, professional but direct, no filler phrases like 'I hope this email finds you well.' Flag any that require a decision from me before replying. For the ARCHIVE candidates: list them in a bulk-archive batch I can process in one click. Do not send anything. All drafts are for my review first."

Before: 600 unread. Start processing. Open email 1: newsletter. Open email 2: automated receipt. Open email 3: client question from three weeks ago, now too late to reply without an apology. Fifty emails in, the energy is gone. 550 still unread. Tomorrow the number will be 570.

After: 600 emails categorized: 12 URGENT, 28 ACTION, 60 FYI, 500 ARCHIVE. 40 reply drafts produced in your voice, ready to review. 500-email bulk-archive batch ready to execute. Total time: 90 minutes to reach inbox zero, including reviewing and sending all 40 replies. Count: 0.

Setup: 5 minutes. Works with Gmail exports, Outlook exports, or pasted email text. Nothing is sent automatically.

2. Smart Email Triage: An Automated Pipeline for the Daily Firehose

This skill builds an automated Gmail classification pipeline with rules for newsletters, receipts, notifications, and action-required emails, automatically labeling and archiving low-priority categories, surfacing a priority inbox that contains only emails needing human response, and generating a daily digest of what was archived so nothing important slips through.

The ongoing triage problem is different from the backlog problem. Once the backlog is cleared, the challenge is keeping it clear. The average professional receives 120 emails per day. Of those, perhaps 15 to 20 require an actual human response. The other 100 are newsletters, automated notifications, receipts, internal announcements, and CC'd threads where no action is required. Manually sorting 100 low-priority emails per day to find 20 that need attention costs 30 to 45 minutes of daily cognitive overhead.

This skill moves that sorting to the system layer. Newsletters get labeled and archived before they ever appear in the inbox. Receipts go to a receipts folder automatically. CC-only threads get tagged for reference without creating inbox pressure. The priority inbox that remains contains only emails that need a human decision.

"Set up a smart email triage pipeline for my Gmail. Rules I want automated: newsletters and marketing to auto-archive with label 'newsletters,' receipts and order confirmations to label 'receipts' and archive, automated notifications (GitHub, Jira, Slack digests) to label 'notifications' and archive, anything where I'm CC'd (not TO) to label 'FYI' and skip inbox, internal company announcements to label 'announcements' and archive. VIP exceptions: any email from [my boss's email] or [key clients] always goes to priority inbox regardless of other rules. Generate a daily digest of everything archived."

Before: 120 emails per day arrive in one pile. First 45 minutes of every morning: manually sorting newsletters, receipts, and Jira notifications to find the 15 emails that actually need a response. Interruptions throughout the day every time a notification email creates false urgency. Inbox is never actually empty.

After: Priority inbox: 18 emails, all requiring human response. 102 others automatically sorted to labeled folders. Daily digest at 5pm: "Archived today: 23 newsletters, 14 receipts, 41 notifications, 24 announcements. Nothing flagged as likely misclassified." Morning sorting time: zero minutes. Inbox zero maintained daily without effort.

Setup: 30 minutes (requires Gmail integration). The VIP exception list is the critical configuration.

3. Email Classifier: Sort Any Volume by Type, Priority, and Required Action

This skill categorizes emails by urgency, required action type, and sender relationship, producing a structured view of any inbox batch that shows exactly which emails need a decision today, which can wait, and which require no action at all.

The classification problem at scale: when 500 emails arrive in a week, the manual sorting process requires the same cognitive decision for each one. Is this urgent? Does this need a reply? Is this from a client, a vendor, or an automated system? That micro-decision repeated 500 times is the hidden productivity tax that email imposes. This skill automates the decision layer, applying consistent rules for urgency and action-type classification across any volume of mail without decision fatigue.

"Classify the following 200 emails. For each email, assign: (1) sender type: client, internal team, vendor/partner, prospect, newsletter, or automated, (2) urgency: urgent (same-day), this-week, or no-action, (3) required action: reply, decision, read-only, or archive. Then give me: a summary table sorted by urgency within each sender type, the total count per category, any emails where urgency is unclear and I need to make a judgment call, and a suggested processing order for the 20 highest-priority items."

Before: 200 emails, all at the same visual priority in the inbox. Scroll through looking for client names. Open anything with an urgent-sounding subject line. Half of them turn out to be newsletter subject lines written to seem urgent. Spend two hours processing 50 emails while the other 150 wait and anxiety compounds.

After: 200 emails classified: 8 urgent-client (reply today), 14 this-week-client, 6 urgent-internal, 22 this-week-internal, 150 archive. Suggested processing order starts with the 8 client urgencies. Processing time for the 50 action items: 45 minutes. The 150 archive items: one batch action. Done by 10am.

Setup: 10 minutes. Works for any email client. No API integration required, just paste or export the email batch.

4. Email Drafter: Write the Hard Emails in 30 Seconds Instead of 20 Minutes

This skill drafts professional emails for any business scenario (client escalations, scope creep pushback, difficult feedback, follow-ups after missed deadlines) calibrated to the context, the relationship, and the tone that the situation requires. Draft ready in under a minute. Review and send in under two.

The drafting friction problem: some emails are genuinely hard to write not because the content is complex but because the tone must be precisely calibrated. An email pushing back on scope creep with a major client must be firm enough to hold the boundary without being adversarial enough to damage the relationship. An email delivering disappointing news to a team member must be honest without being demoralizing. These calibrations are the reason an email that contains 100 words of actual content takes 20 minutes to write.

"Draft an email to a client who is our second-largest account. Context: they have added three features to a fixed-price project that were not in the original scope. We're now 40% over the estimated hours. I need to: flag the scope creep clearly, propose two options (absorb the overrun in exchange for a contract extension, or add a change order for the additional work), and keep the relationship strong. They have a second project starting next quarter. Tone: direct and professional, not apologetic but not adversarial. Do not use filler phrases. I want a clear ask at the end of the email."

Before: Scope creep email: blank draft, 20 minutes of staring, three deleted openings, two versions that sound too aggressive, one version that is too apologetic and does not actually raise the issue clearly. Sent four days late. Client is now confused because the delay made it seem less urgent.

After: Draft produced in 45 seconds: acknowledges the features added, quantifies the hours impact, presents two options clearly, proposes a call to align on the path forward, closes with a positive note about the Q4 project. Reviewed, minor edit to the numbers, sent in 2 minutes. Client responds same day. Change order agreed within the week.

Setup: 10 minutes. Works for any professional email scenario. The more context you provide (relationship, desired outcome, tone constraints), the more precise the draft.

The Email System: Clear, Sort, Classify, Draft

Used in sequence, the workflow for someone starting with an overwhelmed inbox looks like this: the Inbox Zero Manager clears the backlog. The Smart Email Triage is set up to prevent the backlog from returning, sorting incoming mail automatically into priority inbox and labeled archive folders. The Email Classifier handles any batch that accumulates during travel or absence. The Email Drafter handles the replies that would otherwise sit in drafts for days because starting them is too difficult.

The output is not just inbox zero. It is email as a managed system rather than an ambient stressor. The unread count that creates background anxiety becomes a managed queue where every item has a known priority and a known next action. The morning that used to start with 45 minutes of sorting starts instead with a 15-minute review of the 15 to 20 emails that actually need attention. The hard email that used to take 20 minutes to write takes 2. The inbox stops being something to avoid and starts being something to process.

Getting Started

I publish all four email skills as free, downloadable templates at claudecodehq.com: backlog clearance with draft-in-your-voice replies, automated Gmail triage with daily digests, bulk email classification by sender type and urgency, and context-aware email drafting for any professional scenario. Each one is a single file you install once and use whenever you face a backlog, need to sort a batch, or need to write an email that requires more than five minutes. Start with the Inbox Zero Manager on your current backlog. The 90 minutes it takes to reach zero will be the most productive 90 minutes of your week.

Originally published on claudecodehq.com

I Set the Same Goals Every January. This Year I Actually Hit Them. Here's the 4-Skill System.

Daniel Marin — Thu, 25 Jun 2026 17:30:47 +0000

A 30-day protocol that rewires identity, quarterly tracking with evidence-based scoring, an annual audit that replaces vibes with reality, and a personal OS that connects it all.

The resolution failure pattern runs on a reliable schedule. January: goals set, usually the same ones as last year. February: routine disrupted by one bad week, momentum lost. March through November: vague awareness that the goals exist but no active engagement with them. December: reflection that the year drifted, that things did not change the way they were supposed to, that next year will be different.

The standard diagnosis is a willpower problem. The standard prescription is more motivation, a better morning routine, a productivity book. But willpower is not the variable that determines whether goals get achieved. Structure is. A goal without a daily system degrades into an intention. An intention without tracking becomes invisible. A year without an honest audit produces resolutions based on vibes rather than on an accurate picture of where things actually stand.

The Follow-Through Gap

Most goal-setting frameworks focus on the goal itself: what it is, how to state it in measurable terms, how to break it into milestones. These are useful, but they address the wrong bottleneck. The failure point is not goal clarity. Most people can describe their goals clearly. The failure point is the gap between the goal and daily behavior.

That gap has three distinct failure modes. The first is behavioral: no system for daily action means the goal exists only as a mental note, and mental notes fade. The second is tracking: without evidence of progress, momentum is invisible, and invisible progress feels like no progress. The third is identity: a goal that conflicts with your self-image ("I'm not the kind of person who exercises consistently") will be abandoned as soon as friction appears, because the behavior feels foreign rather than natural.

The four skills below each address one layer of the follow-through gap.

1. 30-Day Goal Achievement Protocol: Rewire the Patterns, Not Just the Plan

This skill builds a personalized protocol that addresses the identity layer of goal failure: auditing the self-image and limiting beliefs that cause a behavior to feel unnatural, then building daily identity reprogramming exercises and habit stacking routines calibrated to the specific goal and the specific person trying to achieve it.

The identity mechanism: you cannot sustain a behavior that conflicts with your self-image. Someone who sees themselves as "not a morning person" will abandon a 6am workout habit the moment a bad week provides permission. Someone who sees themselves as "not a finisher" will find reasons to stop a project at 80%. The behavior and the identity are in conflict, and identity wins. Not because of willpower failure, but because the brain is running a coherent self-model and the behavior does not fit it.

The protocol audits the internal operating system first: the self-image around the goal, the specific limiting beliefs that have caused previous attempts to stall, the emotional patterns that trigger avoidance. Then it builds a 30-day daily routine that works at both levels simultaneously: practical habit stacking for the behavioral layer, and daily identity exercises that gradually shift the self-model.

"Build a 30-day goal achievement protocol for me. My goal: launch a side project I've been putting off for 18 months, a Notion template business. Current blockers: I tell myself I'm not technical enough, I'm too busy, and that the market is too saturated. Past pattern: I start strong for 2 weeks, then stop when work gets busy. Give me: an audit of the limiting beliefs in my blockers, a daily identity exercise to shift my self-image from 'person who wants to launch' to 'person who builds and ships,' a weekly milestone structure with checkpoints, a habit-stacking plan I can attach to existing routines, and a trigger-response plan for when work busyness appears."

Before: Notion template business: month 19 of "I'll start next month." New productivity book purchased. New goal-setting framework tried. Strong start for 10 days, then a busy work week, then nothing. The goal is still there. The self-image is still "person who wants to but doesn't."

After: Limiting beliefs audited: "not technical enough" is a fear of judgment, not a skill gap. Daily identity exercise: 2-minute pre-work affirmation tied to first coffee. Week 1 milestone: one template outlined. Work-busy trigger plan: on high-load days, 20 minutes is enough. The streak matters more than the output. Launched in week 6.

Setup: 30 minutes. Works for any goal where previous attempts stalled due to self-image conflict or avoidance patterns.

2. Quarterly Goals Tracker: Evidence-Based Progress Scoring

This skill runs biweekly goal reviews with evidence-based progress scoring (0.0 to 1.0), deadline alerts, and a quarter-end honest assessment, replacing "I think we made progress" with a documented record of exactly where each goal stands and why.

The tracking failure mode: goals set at the start of a quarter are rarely revisited until the quarter ends. By then, the memory of what was intended has faded, the evidence of what happened is scattered, and the assessment is forced to be impressionistic. "I made progress on the fitness goal" means something different than "I hit 9 of 12 weekly workout targets and increased my lifts on two compound movements." The first is a feeling. The second is a record.

The skill scores each goal on a 0.0 to 1.0 scale based on actual evidence, not feeling. A 0.7 on the fitness goal means the evidence supports 70% of the target being met. A 0.3 means 30%, and the review prompts why: was the goal wrong, was the behavior insufficient, were there external factors that should be accounted for? The biweekly cadence means goals are never 10 weeks invisible.

"Set up quarterly goal tracking for Q3. My five objectives: (1) Ship the MVP of my side project by August 31, (2) Reach 185lbs body weight by September 30, (3) Read 6 books this quarter, (4) Have 3 meaningful conversations per week with my team about their development, (5) Run my first 10K by September 15. For each objective: create a progress scoring rubric (what 0.0, 0.25, 0.5, 0.75, and 1.0 look like in evidence terms). Set biweekly check-in reminders. Flag any goals where the deadline creates a sequencing problem. At quarter-end, generate an honest assessment with lessons for Q4."

Before: Five Q3 goals set in June. Next time they are reviewed: late September. "I think I made progress on most of them." The MVP is at 40% but that feels like 70% because the hardest parts are done. The 10K was missed. The team conversations happened "sometimes." Quarter closes on impressions, not evidence.

After: Mid-July check-in: MVP at 0.3 (behind), fitness at 0.6 (on track), reading at 0.5 (on track), team conversations at 0.2 (calendar blocking needed). 10K flagged: 6-week training plan started now barely reaches September 15. Revise to October. Quarter ends with a documented 0.65 average and a Q4 plan that addresses the actual gaps.

Setup: 15 minutes. Works for personal goals, OKRs, and professional development objectives.

3. Annual Life Audit: Honest Assessment Before Direction

This skill produces an honest assessment across every domain of life (career, health, finances, relationships, personal growth), identifies patterns and connections between domains, and breaks the resulting focus areas into quarterly milestones and monthly actions.

The "resolutions based on vibes" problem: most people set annual goals without an accurate picture of where they currently stand. They set a health goal because they feel vaguely unhealthy, not because they have assessed what specifically is working and what is not. The resolution is based on feeling rather than data, so it addresses the wrong thing.

The audit forces the honest assessment first. Across each domain, it asks: what went well, what did not, what surprised you, what are you avoiding looking at? The pattern analysis step is the most valuable output: it surfaces the connections between domains that individual goal-setting misses. Sleep affecting career performance. Financial stress affecting relationship quality. The insight that the same avoidance pattern appears in the health domain and the creative work domain.

"Run an annual life audit for me. I'll answer honestly for each domain. Career: [your responses]. Health: [your responses]. Finances: [your responses]. Relationships: [your responses]. Personal growth/learning: [your responses]. Energy and wellbeing: [your responses]. After I've answered, give me: a summary of what the data says (not what I said I want, what the answers reveal), the 2 to 3 cross-domain patterns you notice, the domain where the biggest gap exists between where I am and where I want to be, and a 3-quarter plan with specific monthly actions starting next month."

Before: December 31: "This year I'll get fit, save more money, and spend more time with family." Same goals as last year. Based on the feeling that those three things are lacking. No assessment of what happened this year. No understanding of why the same goals keep appearing.

After: Audit reveals: the financial stress is driving relationship tension, not "not spending enough time." Career is actually strong. The vague dissatisfaction is about creative work, not performance. Cross-domain pattern: the weeks with poor sleep affect career output, financial discipline, and presence at home. One intervention (sleep) affects three domains. That is the focus area for Q1.

Setup: 10 minutes. Most useful done annually in December or January, but works at any major transition point.

4. Personal OS: One System for Habits, Tasks, Goals, and Reviews

This skill builds a markdown-based personal operating system with daily logs, habit streak tracking, fitness data integration, task management with priorities, and automated weekly reviews, replacing the multi-app fragmentation problem with a single system that makes progress visible across every dimension of life.

The multi-app fragmentation problem: habits tracked in one app, tasks in another, fitness data on a watch that does not export anywhere useful, daily reflections in a journal opened three times this month, goals in a note last edited in January. Nothing talks to anything else. Progress across dimensions (the pattern that poor sleep weeks correlate with missed habit streaks and lower workout output) is invisible because no single system holds all the data.

The Personal OS consolidates everything into markdown files. Daily logs capture mood, energy, tasks completed, and habit check-ins in a consistent format. Weekly reviews run automatically: habit streaks calculated, workout trends analyzed, task completion rates summarized. Because it is markdown, it works with Obsidian, any text editor, or just a folder of files. No app subscription. No data locked in a proprietary format.

"Build me a personal OS for tracking my life. I want to track: daily habits (workout, meditation, reading, no alcohol), fitness data (weight, strength benchmarks), tasks by priority, mood and energy (1 to 5 scale), and weekly goals. Set up: a daily log template I fill in each morning and evening, a weekly review that auto-calculates habit streaks and surfaces what I hit and missed, a monthly trend analysis showing correlations (e.g. do low-energy weeks correspond to missed workouts?), and a simple CLI command to open today's log. Everything in markdown, compatible with Obsidian."

Before: Habits: Streaks app, 12-day streak broken in March, never restarted. Tasks: Todoist, 847 items, last organized in Q1. Fitness: Apple Watch data, never exported. Reflections: Day One, 4 entries this year. Five apps, zero synthesis.

After: One markdown folder. Morning log: 3 minutes. Evening check-in: 2 minutes. Weekly review generated automatically: workout streak 18 days, meditation streak 12, reading 4/7 days, energy averaged 3.4 this week vs 3.8 last week. Monthly trend: low-energy weeks correlate with 2+ missed workouts. The causality is now visible. All in Obsidian, zero app dependency.

Setup: 10 minutes. Works as a standalone system or inside an existing Obsidian vault.

The Goal Achievement Stack

Used in sequence, the workflow looks like this: the Annual Life Audit runs in December and produces the focus areas for the year ahead. The Quarterly Tracker structures those focus areas into quarterly objectives with scoring rubrics. The 30-Day Protocol activates for any specific behavior change embedded in those quarterly objectives, especially ones where previous attempts stalled due to identity conflict. The Personal OS runs daily throughout the year, capturing the data that makes weekly reviews automatic and monthly trend analysis possible.

The pattern that emerges over time is a compounding one. The first year, the audit reveals the real problems rather than the felt ones. The quarterly tracking produces honest assessments rather than rationalized impressions. The daily system surfaces correlations that would otherwise be invisible. By the second year, the goals being set are more accurate, the tracking is more honest, and the patterns that determine whether a behavior sticks are well understood.

The resolution failure pattern breaks not through more motivation but through a system that makes reality legible.

Getting Started

I publish all four goal-setting skills as free, downloadable templates at claudecodehq.com: 30-day identity-based goal protocols, quarterly evidence-based tracking, annual life auditing with cross-domain pattern analysis, and a markdown-based personal OS with automated weekly reviews. Each one is a single file you install once and use whenever you need to set direction, track progress, or reset a stalled system. Start with whichever layer of the follow-through gap is most responsible for your goals not sticking.

Originally published on claudecodehq.com

Every Diet Plan I Tried Was Built for Someone Else. So I Built One for Me With AI.

Daniel Marin — Wed, 24 Jun 2026 12:38:40 +0000

A fat-loss planner calibrated to your actual stats and food preferences, a nutrition advisor that remembers your allergies, a mental health journal that spots patterns over time, and a DNA analysis that turns raw genetic data into a doctor-visit agenda.

Generic health advice fails for a specific reason: it is built for an imaginary average person, not you. The calorie calculator that outputs "eat 2,000 calories" does not know your body weight, activity level, or that you work a desk job but train four times a week. The meal plan full of salmon and broccoli does not know you hate fish and that the only vegetables you will actually eat are the ones in a curry. The journaling app does not know that your anxiety spikes on Sunday evenings before the work week. The DNA test result sits in a zip file with 700,000 genetic markers and no way to make sense of them.

The personalization gap is what makes most health systems fail. The plan that requires willpower to follow is a plan that did not account for the person following it. When the food is food you actually enjoy, following the plan is not a discipline problem. When the journal prompts are calibrated to your situation, journaling becomes useful instead of performative. When the nutrition advice remembers your allergies, you stop having to re-explain yourself every session.

Four AI skills address four distinct health domains: building a nutrition plan calibrated to your actual data and food preferences, maintaining a persistent nutritional advisor that accumulates context across sessions, journaling for mental health with pattern recognition across entries, and analyzing your DNA data for actionable personalized health insights.

The Personalization Gap in Health Apps

Every fitness app promises personalization. Most deliver a slightly customized version of the same generic plan. The calorie target gets adjusted for body weight. The macro split defaults to 40/30/30. The meal plan pulls from a database of 500 recipes and gives you the ones rated highest by users with vaguely similar goals. The foods you hate appear on the Tuesday dinner slot. The workout you cannot do because of a shoulder injury is listed for Wednesday.

Real personalization requires context that most apps cannot hold. They do not know that you find cardio miserable but enjoy lifting. They do not know that you are lactose intolerant and tried to tell the app but it keeps suggesting Greek yogurt. They do not know that your worst week for eating is always the week of a work deadline, and that knowing this pattern might be more useful than a macro target.

AI skills hold context and accumulate specificity over time. The nutrition plan is built from your actual stats and the foods you named. The dietary advisor remembers every preference and restriction you have stated. The journal tracks your entries across sessions and surfaces patterns you would not notice from inside any single day. The DNA analysis runs on your actual genetic data, not population averages.

1. Fat Loss Nutrition Planner: A Plan Built for Your Stats and Your Food

This skill interviews you about your stats, lifestyle, and food preferences, then builds a complete fat loss plan: accurate calorie targets using the Mifflin-St Jeor formula, a macro split calibrated to preserve muscle during the cut, a 7-day themed meal plan built around foods you actually enjoy, snack swaps, a hydration target, and a realistic month-by-month projection.

The generic calculator problem: most online calorie tools use the Harris-Benedict formula with a generic activity multiplier and output a round-number target that is frequently off by 200 to 400 calories in either direction. More importantly, they stop there. They do not produce a meal plan. They do not account for the fact that you love curries and pasta and will not eat a plan that tries to replace them with chicken breast and steamed vegetables.

"Build a fat loss nutrition plan for me. My stats: 34F, 78kg, 165cm, moderately active desk job, train 3x per week (weights + one run). Goal: lose 8kg over 4 months without losing muscle. Foods I love: Indian food, pasta dishes, stir-fries, eggs, Greek food. Foods I hate: fish, brussels sprouts, anything with beetroot. Snacks I reach for: crisps, chocolate, fruit. Give me: exact calorie target using Mifflin-St Jeor, macro split (prioritize protein), a themed 7-day meal plan using foods I listed, smarter swaps for my go-to snacks, hydration target, and a 4-month projection showing expected weekly progress."

Before: Online calculator outputs "1,800 calories." No meal plan. Google "1800 calorie meal plan" and get results full of salmon, cottage cheese, and foods that will not last a week. Stick to it for five days, then abandon it when Tuesday's meal plan requires 45 minutes to prepare after a 10-hour work day.

After: 1,720 kcal target with 145g protein, 170g carbs, 52g fat. Themed week: Egg Tuesday, Stir-Fry Wednesday, Greek Thursday, Indian Friday. Snack swaps: protein bar instead of chocolate, lightly salted rice cakes instead of crisps. Month-by-month projection to target weight. Still eating curry. Still on track after six weeks.

Setup: 5 minutes. Works for fat loss, muscle gain, or body recomposition goals.

2. Nutritional Specialist: Persistent Dietary Advisor That Remembers You

This skill maintains a persistent database of your food preferences, allergies, dietary restrictions, and health goals, applying that accumulated context to every nutrition question and meal planning session without requiring you to re-explain your situation each time.

The context-loss problem is the defining frustration of nutrition apps. You explain that you are lactose intolerant when you sign up. Three weeks later the app suggests a meal plan with Greek yogurt. You correct it. Next month it happens again. The app does not accumulate understanding. It runs each query cold, producing advice for a generic user that gets partially filtered by the preferences in your profile, which only captures what the app asked you on signup.

This skill builds a dietary profile that grows with each session. The first time you mention that you train in the morning and need pre-workout meals that are easy on digestion, that becomes part of the profile. When you mention that your partner is vegetarian and you cook together three nights a week, that gets added. Every session compounds on the last.

"Set up my nutritional profile. Key details: lactose intolerant (severe), trying to build muscle at 80kg body weight, hate cooking more than 30 minutes on weekdays, partner is vegetarian so I cook meat-free 3 nights a week, train 5am before work so need quick pre-workout fuel. Current goal: 160g protein/day. Now give me: a weekday breakfast rotation that is dairy-free, high-protein, and takes under 10 minutes, three quick high-protein dinners I can cook in 30 minutes that work for a vegetarian partner, and pre-workout snack options for 4:30am."

Before: Fifth different nutrition app, fifth time explaining lactose intolerance, fifth time the app suggests a post-workout protein shake that contains whey. The app does not know about the vegetarian partner, the 5am training, or the weekday time constraint. Every session starts from scratch.

After: Dairy-free breakfast rotation: egg scrambles, overnight oats with plant milk, tofu scramble with nutritional yeast. Three 30-minute meat-free dinners hitting 40g+ protein per serving. Pre-workout: banana with peanut butter or a rice cake with almond butter. Lactose intolerance never mentioned again. Profile remembered across every future session.

Setup: 5 minutes. The more sessions you have, the more precise the advice.

3. Mental Health Journal: Guided Reflection and Pattern Tracking Over Time

This skill provides tailored journaling prompts, reflects back what you share, identifies recurring patterns across entries over time, and offers cognitive reframing suggestions, making consistent journaling useful rather than performative.

The blank page problem: most people who want to journal consistently fail not because they lack insight or things to say, but because starting with a blank page produces the same anxious thought loop they were trying to process. "I'm stressed about work" gets written, then nothing follows because there are no prompts to go deeper, no structure to organize the thoughts, and no reflection to make the exercise feel worth the time.

This skill changes the journaling experience from a blank page into a guided conversation. It starts with a prompt calibrated to what you said you want to work through. It asks follow-up questions that pull the surface observation toward underlying patterns. Over multiple sessions, it tracks themes across entries, noticing, for example, that your anxiety consistently spikes in the 48 hours before a particular type of meeting, or that your energy and mood are consistently lower on days when sleep was under six hours.

"I want to start journaling about work stress and my tendency to procrastinate on important projects. Today was frustrating. I had a presentation due and spent most of the day doing low-priority tasks instead. Give me a few prompts to start journaling about this, then after I respond, help me identify what might be underneath the avoidance pattern. I want to journal consistently over the next month. Track the themes that come up across my entries and surface any patterns you notice developing over time."

Before: Open journal app. Blank page. Write "stressed about work." Stare at it. Write "I need to do better with time management." Close app. Same entry as last Tuesday. Nothing processed, nothing changed.

After: Prompted: "What specifically was at stake in the presentation that made it feel easier to avoid?" Reflection after response: avoidance correlates with fear of judgment, not difficulty of task. After three weeks, pattern surfaced: procrastination spikes when the audience for the work includes a specific colleague. Actionable. Not just "I procrastinate."

Important: this skill is a journaling tool, not a therapist or mental health professional. For serious mental health concerns, please reach out to a licensed professional. Crisis resources: 988 (Suicide and Crisis Lifeline), 741741 (Crisis Text Line, text HOME), or your healthcare provider.

Setup: 5 minutes. Most valuable used consistently over weeks.

4. DNA Health Analysis: Turn Raw Genetic Data Into Personalized Insights

This skill takes the raw data file from 23andMe, AncestryDNA, or similar services and searches for health-related genetic variants, surfacing key SNPs across nutrition, metabolism, cardiovascular health, and other domains with context for what each variant means and specific questions to bring to a doctor.

The $99 problem: most people who take a consumer DNA test get their ancestry breakdown, some relative matches, and a raw data file with 700,000 genetic markers they cannot read. The health reports cost extra. Third-party interpretation tools require uploading genetic data to servers with unclear privacy policies. The raw file sits in a Downloads folder doing nothing.

This skill processes the raw data file locally (no upload to a third-party service) and searches for specific SNPs associated with health-relevant traits: MTHFR variants affecting how the body processes folate, APOE variants relevant to cardiovascular and cognitive health, metabolic variants affecting caffeine and alcohol processing, and others. The output is not a diagnosis. It is a personalized research starting point and a doctor visit agenda that the raw data file never was.

"Analyze my 23andMe raw data file for health-relevant genetic variants. [attach raw data file] Focus on: MTHFR variants (folate processing), APOE variants (cardiovascular and cognitive health), BRCA variants (cancer risk awareness), metabolic variants (caffeine processing, alcohol metabolism, vitamin D absorption). For each variant found: what the SNP is, what allele I carry, what the research says about this variant's implications, how it might be relevant to nutrition or lifestyle choices, and specific questions I should bring to my doctor. Format as a clear report I can take to an appointment."

Before: 23andMe raw data file: 700,000 rows of rsIDs, chromosome positions, and genotypes. Completely unreadable. Health reports require an additional purchase. File has been in Downloads for 18 months. $99 test has produced zero health insight.

After: MTHFR C677T: heterozygous, slightly reduced folate processing efficiency, worth discussing methylated folate supplementation with doctor. APOE: typical variant, no elevated cardiovascular or Alzheimer's risk signal. Caffeine metabolism: fast metabolizer variant. Vitamin D: variant associated with reduced synthesis, check D levels at next blood panel. Doctor visit agenda ready.

Important: genetic analysis by an AI is not a medical diagnosis. Variants identified should be discussed with a qualified healthcare provider who can interpret them in the context of your full health picture.

Setup: 15 minutes. Works with raw data files from 23andMe, AncestryDNA, MyHeritage, and other services. Processing happens locally.

The Health Stack

The four skills address four dimensions of the personalization gap.

Fat Loss Nutrition Planner solves the "plan that does not fit my food" problem: building a calorie and macro target from real stats and a meal plan from foods you actually enjoy.
Nutritional Specialist solves the "re-explain my allergies every time" problem: accumulating a persistent dietary profile that makes every future nutrition conversation more precise.
Mental Health Journal solves the blank-page problem: providing structured prompts and cross-session pattern recognition that makes journaling useful instead of performative.
DNA Health Analysis solves the "700,000 unreadable markers" problem: turning a raw data file into an actionable, doctor-ready health report.

Used together, these four skills create a health system calibrated to the specific person, not a population average. The nutrition plan accounts for the food preferences the nutritional specialist has accumulated. The journal patterns connect to the nutrition and energy data from the other skills. The DNA analysis informs the nutritional approach. Each skill reinforces the others because they share the same underlying context: a detailed, accumulated understanding of the specific person using them.

What the skills do not replace is medical care and professional judgment. They are tools for personalized planning and self-knowledge, not replacements for doctors, dietitians, or mental health professionals. The DNA analysis produces questions for a doctor visit, not a diagnosis. The mental health journal supports reflection, not therapy. The nutrition plan is a starting point that a registered dietitian could refine.

Getting Started

I publish all four health and fitness skills as free, downloadable templates at claudecodehq.com: fat loss nutrition planning with themed meal plans, persistent nutritional advising, guided mental health journaling with pattern tracking, and local DNA health analysis. Each one is a single file you install once and configure to your specific details. Start with the Fat Loss Nutrition Planner if you want to see results this month, or the Nutritional Specialist if you want a dietary advisor that finally remembers who you are.

Originally published on claudecodehq.com

I Found $187/Month in Forgotten Subscriptions. Here Are the 4 AI Skills That Fixed My Finances.

Daniel Marin — Tue, 23 Jun 2026 18:14:42 +0000

Budget analysis, expense tracking, multi-agent financial planning, and private statement analysis. How to take control of your money without giving a third-party app access to your bank account.

Most people have a vague sense of their spending. They know they eat out too much, that there are subscriptions they forgot about, that the month always seems to cost more than planned. The bank statement arrives and the total is bigger than expected. Again. But without a precise breakdown, nothing changes. The next month starts with the same vague intention to spend less, no system for tracking whether that is happening, and no plan for where the money should actually go.

The personal finance app landscape offers dozens of solutions (Mint, YNAB, Monarch, Copilot), but they all require connecting a bank account. Giving a third-party app persistent read access to a bank account is a reasonable thing to be uncomfortable with, and it means the people most careful with their money are also the ones least willing to use the tools designed for them.

AI skills work differently. They process the files you already have (CSV exports from your bank, photos of receipts, statements downloaded from your account page) and run the analysis locally. No bank login required. No data uploaded to a third-party server. Four skills cover the full personal finance problem: understanding current spending, tracking ongoing expenses, planning the financial future, and analyzing statements privately.

The Money Blindspot Problem

The reason most people fail at budgeting is not willpower. It is information. A budget built on estimates ("I probably spend about $400 on food") fails the moment real spending is $620. A budget built on last month's actual categories fails because the category that ran over is always a different one this month. And a budget that requires manual entry of every transaction fails within two weeks because manual entry is tedious and the data is always behind.

The specific version of this problem that money apps do not solve: subscription creep. The $12.99 streaming service from 2021 that was supposed to be temporary. The $9.99 cloud storage upgrade. The $14.99 app subscription that was free for the first year. Each one is small enough to ignore. Together they add up to a number that would be alarming if anyone added it up. Most people do not add it up because doing so requires going through twelve months of bank statements and matching recurring charges. Work that takes hours and that nobody does voluntarily.

1. Budget Analyzer: Find Where the Money Actually Goes

This skill takes bank statements or transaction exports and builds a clear picture of actual spending: every transaction categorized, patterns identified by category over time, subscription creep surfaced, and a realistic budget derived from real habits rather than aspirational estimates.

The core insight: a budget based on what you think you spend will always be wrong. A budget based on what you actually spent last quarter is calibrated to reality. The skill does the categorization work that makes the difference, grouping hundreds of transactions into housing, food, transport, subscriptions, entertainment, and miscellaneous, then showing the totals in a way that makes the surprise spending categories obvious.

"Analyze my last 3 months of bank statements. [paste CSV export] Categorize every transaction into: housing, groceries, dining out, transport, subscriptions, entertainment, health, shopping, and miscellaneous. Then give me: total spent per category per month, which categories exceeded $500/month, every recurring subscription charge with monthly cost and total over 3 months, the 5 highest individual transactions, and a realistic monthly budget for each category based on my actual patterns. Not ideal targets."

Before: $3,200 left the account in October. Dining out "maybe $400." Subscriptions "probably $80." Actual grocery spend unknown. It blurs with pharmacy and Target runs. Budget spreadsheet has been at "85% complete" for six weeks. Month closes, nothing changes.

After: Dining out: $683. Subscriptions: $187/month across 14 services, 6 of which haven't been used in 90 days. Groceries vs. pharmacy separated. A realistic monthly budget built from 3 months of actuals, not guesses. Four subscriptions cancelled same day: $76/month recovered.

Setup: 5 minutes. Works with any bank's CSV export. No bank login required. The subscription audit alone typically recovers more than enough to justify the setup time.

2. Expense Tracker: Keep the Ledger Current Without Manual Entry

This skill processes receipts, bank exports, and transaction records into a running ledger with automatic categorization, approval workflows for business expenses, and monthly summary reports, replacing the end-of-month reconstruction problem with a system that stays current.

The expense tracking problem has two distinct versions. For individuals and freelancers, it is the receipt chaos problem: photos taken on the phone, paper receipts stuffed in a wallet, digital receipts buried in email. By the time tax season arrives, reconstructing the year's expenses from scattered sources is a multi-day project that produces approximate answers at best. For small business owners and teams, it is the approval and reimbursement problem: expenses submitted late, categorized wrong, waiting in someone's inbox for approval, and reconciled weeks after the fact.

"I'm a freelancer who needs to track all business expenses for quarterly tax filing. Categories: software/subscriptions, home office (25% of utilities and rent), professional development, equipment, client meals, and travel. Process these receipts and bank transactions from Q2: [paste data]. Give me: a categorized ledger with date, vendor, amount, and category for each item, total per category for the quarter, items that need receipts I haven't provided, and a summary formatted for my accountant."

Before: Q2 tax filing arrives. Receipts are in three places: phone camera roll, email search for "receipt," paper pile on the desk. Three days reconstructing expenses, approximating anything over 60 days old, missing $400 in deductions because the receipts are unfindable. Accountant gets a rough spreadsheet and a lot of "approximately."

After: Categorized Q2 ledger: 87 transactions, $14,300 total, broken down by category with totals. Four items flagged as needing receipts. Home office calculation done automatically at 25%. Summary formatted for accountant handoff. Filed on time. No approximations.

Setup: 10 minutes. Works for individuals tracking personal deductions, freelancers filing quarterly taxes, and small teams managing reimbursement workflows.

3. Multi-Agent Financial Planner: Four Specialists Working Your Numbers

This skill builds a virtual financial advisory team (a budget analyst, a retirement planner, an investment advisor, and a coordinating lead) that analyze your financial data from different angles simultaneously and produce a unified plan with specific recommendations.

The single-advisor problem: financial planning involves budgeting, investing, taxes, and retirement simultaneously, and these dimensions interact in ways that single-angle analysis misses. A budget advisor who does not know your retirement shortfall will tell you to save more without knowing how urgently. A retirement planner who does not know your tax situation will miss optimization opportunities. An investment advisor who does not know your cash flow will recommend allocations you cannot sustain.

The multi-agent approach runs all four analyses concurrently and then synthesizes them. The lead partner synthesizes all four into a single action plan with quarterly milestones. The kind of comprehensive picture that previously required hiring multiple specialists at thousands of dollars per year.

"Create a comprehensive financial plan for my household. Here is our data: combined income $145,000/year, monthly expenses $8,200 (housing $2,800, food $900, transport $600, subscriptions $180, childcare $1,400, misc $2,320), savings rate 8%, current retirement accounts $94,000 (60% stocks / 40% bonds), target retirement age 62, current age 38, no employer match. Run four-agent analysis: budget, retirement, investment, and tax. Give me: where we can free up cash, whether we're on track for retirement at 62, whether our investment allocation matches our timeline, what we're missing on taxes, and a 90-day action plan with specific steps."

Before: Annual "financial checkup" with a generic online calculator that says "you need $1.8M to retire." No path to get there. Budget review happens after a bad month, not systematically. Investment allocation was set in 2019 and hasn't been revisited. Tax situation: "I think we're doing fine."

After: Budget agent: $780/month available from childcare-to-activity transition in 18 months. Retirement agent: 4 years behind target at current rate; need $1,100/month additional. Investment agent: 60/40 allocation is 10% too conservative for a 24-year horizon. Tax agent: HSA contribution gap, $3,200 in missed deductions. 90-day action plan with specific account changes.

Setup: 15 minutes. Works for individuals, couples, and families at any income level.

4. Finance Manager: Private Statement Analysis With Charts, Locally

This skill imports bank statements (CSV or PDF) and produces instant spending breakdowns, 50/30/20 budget comparisons, savings recommendations, and interactive charts, with all processing done locally on your machine. Nothing is uploaded. No bank login is required.

The privacy gap in personal finance apps is real. Persistent bank access means the app can pull your transaction history at any time, not just when you explicitly request it. The terms of service for most budgeting apps permit using transaction data for product improvement, marketing, and in some cases sale to third parties. For people who are careful about financial data (which is a reasonable thing to be), this is a dealbreaker that leaves them with no good options: manual spreadsheets or no analysis at all.

This skill changes the calculus. You download the statement yourself from your bank's secure website (which you already trust), import the CSV or PDF, and get the same analysis that the apps provide: spending breakdown by category, comparison against the 50/30/20 rule (50% needs, 30% wants, 20% savings), the top five surprise expenses, and three specific savings recommendations.

"Analyze my Chase bank statement from last month. [paste CSV] Give me: (1) spending breakdown by category with percentage of total income, (2) comparison to the 50/30/20 rule, which buckets am I over or under, (3) top 5 highest individual transactions with notes on whether they're recurring or one-time, (4) 3 specific savings recommendations based on where I'm overspending relative to the 50/30/20 targets, and (5) a simple month-over-month trend I can track going forward. All processing local. Do not summarize or send this data anywhere."

Before: The choice between giving Mint read access to the bank account forever, or downloading a CSV every month and staring at 200 rows of transactions in Excel with no real analysis. Most months: download, wince at the total, close the file. Nothing changes because the data is unprocessed.

After: Spending breakdown with percentages. 50/30/20 comparison: needs 58% (over by 8%), wants 29% (on target), savings 13% (under by 7%). Top 5 transactions: three recurring, two one-time. Three specific recommendations: reduce dining out by $150, pause unused gym membership, redirect to savings. Processed locally, no data exposure.

Setup: 10 minutes. Works with CSV exports from any bank and most PDF statement formats.

The Personal Finance Stack

The four skills address the four stages of the personal finance problem.

Budget Analyzer answers "where did the money go?" for the past: categorizing historical spending and surfacing the subscription creep and category surprises that a vague sense of the numbers misses.
Expense Tracker keeps the current month's ledger up to date: processing receipts and bank exports as they arrive instead of reconstructing them at month-end.
Multi-Agent Financial Planner addresses the future: stress-testing retirement timelines, investment allocations, and tax situations simultaneously with the rigor of four specialized advisors.
Finance Manager ties it together for the privacy-conscious user who wants the analytical depth of a budgeting app without the bank login.

Used together, the workflow looks like this: the Budget Analyzer runs on the last three months of statements to establish the baseline. The Finance Manager runs monthly on each new statement as it arrives, tracking progress against the 50/30/20 targets. The Expense Tracker handles receipts and tax-relevant expenses throughout the year so quarterly filing is not a reconstruction project. The Multi-Agent Financial Planner runs quarterly or after any major financial change to keep the long-term plan current.

What the skills do not replace is financial judgment: the decision about whether to prioritize the emergency fund or pay down debt first, whether the retirement target date is realistic given lifestyle goals, whether a major purchase now is worth the delayed savings impact. Those decisions require the person who knows the full picture. What the skills provide is the full picture, precise, current, and analyzed from multiple angles, so those decisions are made with real information instead of approximations.

Getting Started

I publish all four personal finance skills as free, downloadable templates at claudecodehq.com: budget analysis with subscription auditing, expense tracking with receipt processing, multi-agent financial planning, and private statement analysis with 50/30/20 comparison. Each one is a single file you install once and use whenever you download a bank statement, process receipts, or want to revisit your financial plan. Start with the Budget Analyzer on your last three months of statements. The subscription audit alone will probably pay for your afternoon.

Originally published on claudecodehq.com

I Read 60 Papers and Wrote My Lit Review in a Week. Here Are the 4 AI Skills That Made It Possible.

Daniel Marin — Mon, 22 Jun 2026 20:21:59 +0000

An academic research assistant, a literature review builder, a PDF interrogation tool, and a reading analysis system. How students can study smarter without working harder.

The academic workload problem is not a time problem. It is a throughput problem. There is more to read than time to read it, more papers to survey than hours to screen them, more arguments to synthesize than working memory to hold them. A PhD student writing a dissertation literature review does not have a laziness problem. They have a structural problem: the tools for finding, organizing, synthesizing, and retaining academic material are fragmented (a database search here, a Zotero library there, highlights in a PDF app, notes in a separate document) and none of them talk to each other.

AI skills replace the fragmented toolchain with persistent assistants tuned to the specific tasks that eat academic time: building and executing literature searches, synthesizing reading lists into structured reviews, interrogating individual documents without reading every word, and turning passive reading into connected knowledge.

The four skills below cover the full academic research cycle: finding sources, reviewing the literature, getting answers from specific documents, and building a cross-source knowledge base. Used individually, each saves hours per assignment. Used together, they change what it is possible to accomplish in a semester.

The Academic Throughput Problem

Every student who has written a literature review knows the specific feeling of being buried. The database search returns 847 results. You screen titles and narrow to 200 abstracts. You read abstracts and pull 60 papers. You read 60 papers over three weeks and take notes in a combination of margin scrawl, sticky notes, and a Google Doc that has grown to 47 pages with no structure. Now your advisor asks for a thematic synthesis of the field, organized not by paper but by insight.

The problem is not that you did not read the papers. The problem is that reading without a system for capture, organization, and synthesis produces facts without structure. You remember that three papers used randomized controlled trials, but you cannot immediately reconstruct which three, or how their findings compared, or whether they all agreed. You remember there was something about a gap in longitudinal studies, but you cannot find where you wrote that down.

The four skills below encode the structural layer that most students try to improvise.

1. Academic Research Assistant: Search, Screen, and Synthesize Literature

This skill builds and executes systematic literature searches, screens abstracts for relevance, manages citations, compares methodologies across papers, and identifies gaps in the existing research, transforming weeks of manual literature review work into a structured, repeatable process.

The core problem it solves is not just finding papers. It is knowing which papers matter and why. A database search that returns 847 results is not useful until those results are screened by relevance to a specific research question, sorted by methodology, and assessed for how each contributes to or contradicts the emerging picture of the field. That screening and sorting is the work that takes weeks, and it is the work that gets done inconsistently when left to unstructured note-taking.

The skill maintains a structured database of every paper assessed: relevance score against the research question, methodology classification, key findings, limitations, and theoretical framework. When it is time to write the literature review, the database does the organizing.

"Build a literature search on the effects of sleep deprivation on undergraduate academic performance. Research question: does chronic sleep restriction (less than 6 hours per night) have measurable effects on GPA, test scores, and cognitive performance in college students? Search PubMed, PsycINFO, and Google Scholar. Screen results and return: the 20 most relevant papers with relevance scores (1 to 10), methodology classification (RCT, longitudinal, cross-sectional, meta-analysis), key findings, and limitations. Then identify: the 3 most cited methodological approaches, 2 major contradictions in the literature, and the main research gap my dissertation could address."

Before: Three weeks of database searches, 200 abstracts read, 60 papers pulled, notes scattered across a 47-page Google Doc. No structure. Cannot reconstruct which papers used RCTs. Cannot find where the note about longitudinal gaps was written. Literature review chapter is three months late.

After: 20 papers ranked by relevance score with methodology classifications, key findings table, limitations column, contradiction map, and a research gap statement ready to paste into the dissertation introduction. Literature review chapter drafted in one week instead of three months.

Setup: 10 minutes. Works for any academic field with searchable databases.

2. Literature Review Builder: From Reading List to Thematic Narrative

This skill tracks papers with their methodology, findings, and limitations, groups them by emergent theme, and drafts a narrative literature review organized by insight rather than by source. The difference between a list of summaries and an actual synthesis of the field.

The literature review problem is not reading the papers. It is the step after reading: turning 40 individual paper summaries into a coherent argument about where the field is and where it is going. Most students have a list of summaries in their notes and a blank document where the synthesis should be. The blank document stays blank because synthesis is not "write up each paper in turn." It is "identify the five themes that run through all 40 papers, show how different papers relate to each theme, and build an argument about what the collective literature tells us and what it leaves unaddressed."

The skill does the thematic grouping. Given 40 papers with their methodologies and findings, it identifies the recurring themes, groups papers under each theme, builds the methodology comparison table that advisors ask for, surfaces contradictions between papers that need to be addressed in the review, runs a gap analysis showing what is under-researched, and drafts the narrative organized by theme with proper citations.

"I have 40 papers on remote work productivity. Here are the titles, authors, methodologies, and key findings for each. [paste list] Build a literature review organized by emergent theme. I need: 4 to 6 themes with the papers grouped under each, a methodology comparison table, 3 contradictions between papers that I need to address in my review, a gap analysis identifying what's under-researched, and a narrative draft of the introduction and first two theme sections with citations in APA format."

Before: 87 papers in Zotero. A blank Google Doc titled "Lit Review Draft." The advisor said to organize it thematically, not as a list of summaries. Every attempt to start produces a different structure. Two weeks of staring at the blank document, three failed outlines, deadline in four days.

After: 5 themes identified with papers grouped under each. Methodology comparison table. 3 paper contradictions flagged for explicit discussion. Gap analysis pointing to under-researched area (longitudinal studies post-2020). Narrative draft of introduction and first two theme sections. Review complete in two days.

Setup: 10 minutes. Works best with 15+ papers where thematic synthesis is more valuable than sequential summary.

3. Chat With PDF: Interrogate Any Document Without Reading Every Word

This skill answers questions about PDF content, extracts specific information, and summarizes sections, giving students an interactive interface to interrogate any document rather than reading it cover-to-cover to find one answer.

The dense document problem: a 200-page annual report, a 60-page research paper, a textbook chapter with no clear structure. Ctrl+F only works when you know the exact words the author used. Reading the whole document to find one figure, one clause, or one methodological detail takes an hour. The mental load of holding a long document in working memory while searching for one piece of information is enormous, and often the information turns out to be in a section you skimmed.

This skill gives every document an interface. "What methodology did they use in section 3?" "What are the three main risk factors mentioned?" "Summarize the findings on pages 45 to 60." Each answer comes with page references so the source can be verified.

"I have a 150-page annual report for a company I'm analyzing for my finance class. I need: (1) the three main risk factors they disclose, with page references, (2) the revenue growth figures for the last three fiscal years, (3) a summary of their stated competitive advantages, (4) any mention of pending litigation or regulatory issues, and (5) a one-paragraph executive summary of the business model and current strategic priorities. Cite specific pages for each answer."

Before: Two hours reading a 150-page annual report. The risk factors are somewhere in the first 40 pages, the revenue figures are in the financial statements, and the strategic priorities might be in the CEO letter or in a section titled something different. Ctrl+F for "risk" returns 47 results. Notes are scattered and un-cited.

After: Three risk factors extracted with page references. Revenue growth figures for three years pulled from financial statements. Competitive advantages summarized. Litigation disclosures flagged. One-paragraph executive summary ready to use. Total time: 10 minutes. Every answer verifiable by page number.

Setup: 10 minutes. Works for any PDF: research papers, textbook chapters, legal documents, annual reports, government publications.

4. Reading Analysis System: Turn Passive Reading Into Connected Knowledge

This skill captures each book's key arguments, connects insights across books, surfaces contradictions between authors, identifies gaps in your reading, and recommends what to read next, turning a reading list into a cross-source knowledge base rather than a list of titles that blur together.

The retention problem is structural, not motivational. Reading 30 books in a year produces 30 isolated memories that decay at the same rate as every other experience. Without a system for capturing what each book argued, connecting arguments across books, and identifying where different authors disagree, reading is entertainment, not learning. Most students discover this when asked to explain what they have read: they can name the books, describe the general topic, and recall one or two vivid examples. But they cannot reconstruct the argument, cannot compare it to adjacent books, and cannot explain how their reading has changed what they think.

The reading analysis system changes the unit of capture from "what I remember" to "what the book argued." Each book gets a structured entry: core thesis, supporting arguments, evidence methodology, key examples, limitations the author acknowledges, and how it connects to or contradicts other books already in the system.

"Analyze my reading on behavioral economics over the past year. Books: Thinking Fast and Slow (Kahneman), Nudge (Thaler and Sunstein), Predictably Irrational (Ariely), Misbehaving (Thaler), The Undoing Project (Lewis). For each book: core thesis in 2 sentences, 3 key arguments, methodology (lab experiments, field studies, theory, narrative), main limitation the author acknowledges. Then cross-book synthesis: 3 themes that appear across 3+ books, 2 contradictions between authors where they disagree, what aspect of behavioral economics my reading has not yet covered, and 3 books I should read next to fill that gap with reasons."

Before: Five behavioral economics books read over a year. Can name them all. Can describe what each is "about." Cannot reconstruct the core argument of any of them in two sentences. Cannot explain whether Kahneman and Thaler actually agree or what they disagree about. Reading has not compounded into knowledge. It has evaporated into impressions.

After: Core thesis for each book. Three cross-book themes (cognitive bias persistence, policy intervention design, methodological limits of lab-to-field generalization). Two Kahneman/Thaler contradictions identified. Gap: statistical decision theory and the mathematics underlying behavioral models. Three next reads with reasons. The year of reading is now a knowledge base, not a list.

Setup: 5 minutes. Works for academic reading, professional development books, and any cross-disciplinary reading where connections between sources matter.

The Student Research Stack

The four skills map to the four stages where academic work gets stuck.

Academic Research Assistant handles the front end of any research project: finding the right papers, screening them systematically, and identifying what the field has not addressed.
Literature Review Builder handles the synthesis step: taking a reading list and producing the thematic narrative that advisors actually want to see.
Chat with PDF handles the document interrogation problem: getting specific answers from specific sources without reading cover-to-cover.
Reading Analysis System handles the long-term retention problem: making sure that what gets read actually compounds into knowledge.

For a student writing a dissertation, the workflow looks like this: the research assistant finds and screens the literature, producing a structured database of assessed papers. The literature review builder takes that database and produces the thematic synthesis chapter. Chat with PDF handles the deep dives into individual papers where specific claims need to be verified and page-referenced. The reading analysis system maintains the broader knowledge base, connecting the dissertation reading to everything else the student has been reading.

For a student with coursework rather than a dissertation, the stack simplifies. Chat with PDF handles every dense reading assignment (textbook chapters, primary sources, case studies) by turning sequential reading into targeted interrogation. The reading analysis system handles the synthesis questions that appear on exams and in seminar discussions: not "what did this book say" but "how does this book connect to the others we have read and what does the conversation between them tell us."

What the skills replace is the improvised system: the scattered notes, the unstructured Zotero library, the blank synthesis document, the reading that evaporates. The throughput problem does not go away, but the structural layer that converts throughput into knowledge is no longer something each student has to invent from scratch.

Getting Started

I publish all four student skills as free, downloadable templates at claudecodehq.com: systematic literature search and screening, thematic literature review building, PDF interrogation with page references, and cross-source reading analysis. Each one is a single file you install once and use whenever the task comes up. Start with whichever stage of the research cycle is currently blocking you. For most students, that's either the literature review (if you have papers but no synthesis) or Chat with PDF (if you have a dense document and a deadline).

Originally published on claudecodehq.com

I Built a Full 6-Module Course in One Session. Here Are the 4 AI Skills Every Educator Needs.

Daniel Marin — Sun, 21 Jun 2026 13:52:01 +0000

From curriculum outline to LMS-ready quizzes, personalized syllabi, a self-hosted AI tutor, and vocabulary lists that match what learners actually need.

Every teacher, trainer, and course creator faces the same hidden labor problem: knowing your subject deeply is not the same as knowing how to teach it. Curriculum design, lesson sequencing, assessment creation, vocabulary scaffolding: these are instructional design skills that are entirely separate from subject-matter expertise. A data scientist who knows Python cold still has to figure out how to structure a six-module course, write 24 lesson objectives, create quizzes that test understanding rather than memorization, and produce a capstone project that ties it all together.

Most educators solve this by spending weekends on it. The lesson plan gets built Sunday night. The quiz gets assembled from memory. The course outline gets stared at for two hours before the first module gets written. The vocabulary list for the language class gets pulled from a generic textbook that covers words nobody actually uses in the context the students care about.

AI skills compress the structural work without replacing the teaching judgment. The four skills below cover the full range of educator needs: building a complete course, personalizing a learning path, setting up a tutoring system for students, and creating vocabulary materials that actually match what learners need. The educator's expertise stays at the center. The skills handle the instructional architecture around it.

The Instructional Design Gap

Instructional design is a profession for a reason. The gap between "I know this subject deeply" and "I can teach this subject effectively" is filled by decisions that experienced instructional designers make systematically: what prerequisite knowledge to assume, how to sequence concepts so each one builds on the last, when to introduce a quiz versus when more explanation is needed first, how to write a learning objective that is specific and measurable rather than vague and aspirational.

Most educators learn these decisions by trial and error over years. Corporate trainers build courses by adapting the previous course. Language teachers reuse the same vocabulary lists because creating new ones takes longer than the lesson itself. Subject matter experts moving content to an LMS for the first time face a blank curriculum template and have no model for what goes where or why.

The skills below encode instructional design principles into the structural scaffolding they produce. The educator provides the expertise (the right sequencing for this particular subject, the examples that will land with this particular audience, the assessment questions that test the understanding that actually matters). The skill provides the architecture: the module structure, the lesson framework, the syllabus sequencing logic, the spaced repetition scheduling.

1. E-Learning Course Creator: From Expertise to Full Curriculum

This skill plans, structures, and writes complete e-learning courses through a phased approach: curriculum design, lesson writing, and assessment creation, producing a full course with learning objectives, lesson content, module quizzes, and a capstone project formatted for LMS import.

The skill addresses the specific problem that stalls course creation: the blank curriculum outline. A data scientist who wants to build a Python course for analysts knows the content (NumPy, pandas, data visualization, machine learning basics) but faces a structural question before writing a single lesson: what is the right module sequence, how many lessons per module, what does each lesson need to accomplish, and what quiz questions will test whether the lesson actually landed? These are instructional design questions, not Python questions.

"Create a 6-module Python course for data analysts who know Excel but have never coded. Learning goal: comfortable doing data cleaning, analysis, and visualization in Python by the end. Module topics: Python basics and environment setup, pandas for data manipulation, data cleaning and missing value handling, exploratory data analysis with matplotlib and seaborn, writing reusable functions and scripts, and a capstone project. For each module: 4 lesson titles with learning objectives, 1 worked example using realistic analyst data, and 5 quiz questions with answer keys. Format the capstone as a guided project with evaluation rubric. Structure for Teachable import."

Before: Three weekends staring at a blank curriculum outline. The module sequence keeps changing. Module 3 feels like it needs to come before Module 2. The quiz questions are an afterthought written the night before publishing. The capstone is vague. The course launches six months after the original plan.

After: Full curriculum: 6 modules, 24 lesson titles with behavioral learning objectives, worked examples using realistic analyst data, 30 quiz questions with answer keys and wrong-answer explanations, and a capstone project with rubric. Structured for Teachable import. Course ready to record in one session.

Setup: 10 minutes. Works for technical courses, professional training, corporate learning programs, and hobbyist content. The phased approach (curriculum first, then lessons, then assessments) mirrors how instructional designers actually work.

2. Personalized Syllabus: A Learning Path Built for This Specific Learner

This skill builds a study plan tuned to the learner's existing knowledge and available time: ordered texts or topics, themes to track across the material, explicit bridges from what the learner already knows to what they are about to learn, and an interactive companion mode for working through difficult passages.

The one-syllabus-for-everyone problem is structural. A university course assumes a median student with a particular prerequisite background. But the actual range of students in any class spans from someone for whom the first three weeks are review to someone for whom the first three weeks are completely new territory.

For educators, this skill is most valuable for designing differentiated learning paths. For individual learners, it builds the personal study plan that accounts for what they specifically already know. The explicit bridge between existing knowledge and new material is the skill's most distinctive output: not just "read Chapter 3," but "Chapter 3 introduces the concept of X. You will recognize this as analogous to Y from your background in Z, with the key difference being that..."

"Build a personalized study plan for a marketing manager with 8 years of digital marketing experience who wants to deeply understand data science. Available time: 6 hours per week for 12 weeks. Starting point: comfortable with Excel, basic statistics (mean, median, standard deviation), A/B testing concepts. Give me an ordered curriculum with time allocations, explicit bridges from digital marketing concepts she already knows to the data science concepts she is about to learn, the 3 themes to track across all the material, and recommendations for how to use the interactive companion mode when she hits a concept that doesn't click."

Before: Generic "data science for marketers" course with 40 hours of content, half of which covers statistics she already knows from A/B testing. She gets three weeks in, hits a section on linear algebra that has no connection to anything she knows, loses momentum, and stops. Six months later the course is marked 35% complete.

After: 12-week curriculum skipping what she already knows. Every new concept explicitly bridged: "regression is the mathematical formalization of what you do intuitively when you attribute revenue changes to campaign variables. The model just makes the weighting rigorous." Three themes to track. Companion mode activates when she hits a passage she wants to interrogate. Completes in 11 weeks.

Setup: 10 minutes. Works for any subject area and any learner background.

3. DeepTutor: A Self-Hosted AI Tutoring Platform With Your Own Knowledge Base

This skill sets up a persistent AI tutoring system with a RAG knowledge base built from the educator's own materials (textbooks, notes, PDFs) that generates quizzes, runs Socratic tutoring sessions, and maintains memory across sessions.

The difference between this and a standard "upload a PDF to ChatGPT" approach is persistence and depth. Uploading a textbook to a chat interface gives a Q&A bot that forgets the conversation the moment the window closes. This skill gives a persistent TutorBot with its own knowledge base built from the educator's specific materials (the actual textbook for this course, the lecture notes for this semester, the supplementary papers for this topic) that a student can return to across multiple sessions, and that generates quizzes specifically grounded in that source material rather than in the model's general knowledge.

For educators, this means building a tutoring assistant that teaches from the same materials as the course. A student asking about thermodynamics gets an explanation grounded in the course's textbook chapters, not a different explanation from a different framing that may conflict with how the concept was introduced in lecture.

"Set up a physics tutor for my introductory mechanics course. Course materials: chapters 1 to 8 of University Physics (Young and Freedman), my lecture slides for all 8 chapters, and 3 supplementary problem sets. Build a RAG knowledge base from these materials. Configure a Socratic TutorBot that asks questions before giving answers. When a student asks 'how do I solve this kinematics problem?' the bot should probe what they already know before providing the solution path. Generate a 20-question quiz covering chapters 1 to 4 with difficulty progression."

Before: Students use ChatGPT for homework help. The explanations they get contradict the course notation. The "tutor" has no memory of previous sessions. A student who asked about Newton's second law on Tuesday has to re-explain their confusion on Thursday. Office hours are the only consistent source of course-specific help.

After: TutorBot grounded in the course's own textbook chapters and lecture slides. Socratic mode probes student reasoning before providing solutions. Memory persists across sessions. The bot remembers what the student struggled with on Tuesday and follows up on Thursday. 20-question quiz generated from course-specific material with difficulty progression.

Setup: 15 minutes. Works for any subject with PDF or document-based source materials. The Socratic mode is configurable.

4. Language Vocabulary Builder: Targeted Lists With Spaced Repetition

This skill creates personalized vocabulary lists with context sentences, pronunciation notes, and Anki-ready flashcards with spaced repetition scheduling, targeted to the learner's specific use case rather than pulled from a generic textbook frequency list.

The generic vocabulary list problem: Duolingo teaches "the cat eats bread." A standard frequency list teaches the 2,000 most common words in Spanish, many of which a business professional will never use in their actual work. A traveler preparing for a trip to Japan does not need the 500 most common Japanese words. They need the vocabulary for ordering food, navigating transportation, managing accommodations, and handling unexpected situations. A medical professional learning Spanish needs clinical terminology, not conversational filler.

The skill generates vocabulary lists from the learner's specific context: their job, their travel plans, their exam requirements, their actual conversation needs.

"Build a Spanish vocabulary set for a software engineer who needs to manage a team in Mexico City. Focus areas: giving and receiving technical feedback, running sprint ceremonies (standup, retrospective, planning), discussing system architecture decisions, handling escalations professionally, and casual team relationship-building. 150 words and phrases organized by situation. Each entry needs: the word/phrase, pronunciation guide, English translation, 2 context sentences showing it in a software team context, and any register notes (formal vs. informal). Export as Anki-ready flashcards with a spaced repetition schedule for a 60-day learning plan."

Before: Generic Spanish course teaches conversational vocabulary for restaurants, hotels, and shopping. The engineer learns "Dónde está el baño?" and "Quiero una mesa para dos." In the first standup with the Mexico City team, they have no vocabulary for "blocking issue," "sprint velocity," or "refactor the authentication layer."

After: 150 words and phrases organized by sprint ceremony, architecture discussion, escalation handling, and team relationship-building. "Bloqueado," "velocidad del equipo," and "refactorizar" with context sentences from actual engineering conversations. Anki export with spaced repetition scheduling. Functional in standups within 30 days.

Setup: 5 minutes. Works for any language and any use-case context. Language teachers use it to build differentiated vocabulary sets for different student groups. Learners use it to build their own targeted lists.

The Educator's System

The four skills address different dimensions of the same core challenge: matching the structure of teaching to the needs of learning.

E-Learning Course Creator handles the curriculum architecture when the educator is building from scratch.
Personalized Syllabus handles differentiation when different learners need different paths through the same material.
DeepTutor handles the between-class support layer: the tutoring that previously only happened in office hours, now available whenever the student is studying.
Vocabulary Builder handles materials creation for language and vocabulary work, where generic lists fail learners with specific needs.

What the skills replace is not the educator's expertise or judgment. It is the blank-page problem. The curriculum outline that takes three weekends to write takes one session. The personalized syllabus that would require a one-on-one consultation with every learner gets produced at scale. The tutoring assistant that requires a teaching assistant to build and maintain runs from the educator's own materials. The vocabulary list that takes two hours per topic group takes minutes.

The teaching judgment (what this audience needs to understand, what misconceptions to pre-empt, which examples will land and which will confuse, how to assess whether the learning objective was actually achieved) stays with the educator. The instructional architecture gets handled.

Getting Started

I publish all four education skills as free, downloadable templates at claudecodehq.com: e-learning course creation, personalized syllabus building, self-hosted AI tutoring with RAG knowledge bases, and context-targeted vocabulary building with Anki export. Each one is a single file you install once and use whenever you build a course, design a learning path, or create instructional materials. Start with whichever blank page is staring at you right now.

Originally published on claudecodehq.com

I Apply Minto, PESTEL, and SWOT Faster Than Any Associate I've Trained. Here's How.

Daniel Marin — Fri, 19 Jun 2026 11:18:37 +0000

Four AI skills mapped to the frameworks consultants already use: Minto Pyramid for answer-first logic, PESTEL for rigorous macro scanning, a 10-framework strategy toolkit, and a proposal generator that reflects what the prospect actually said.

Consulting is sold as structured thinking. Clients pay for a framework that organizes their problem, a macro scan that surfaces the forces they are not tracking, a strategy analysis that names the options clearly, and a proposal that demonstrates you actually listened in the discovery call. The frameworks are the product. Not just the output, but the signal of rigor that earns trust before the engagement begins.

The irony is that applying those frameworks takes time that most consultants spend under deadline pressure. The Minto Pyramid does not build itself from a draft that buries the answer in paragraph three. A PESTEL analysis requires filling six cells with specific, sourced intelligence, not vague macroeconomic observations. A strategy session that covers SWOT, go-to-market, pricing, and KPI design thoroughly takes days to prepare from scratch. A proposal that reflects the prospect's actual objections requires re-reading discovery call notes and mapping responses to each concern before writing a single word.

AI skills compress that preparation without changing the intellectual work. Each one is built around a specific named framework that consultants already use. The frameworks run faster. The rigor stays.

Why Consultants Should Be Skeptical of Generic AI

The consulting profession has good reasons to be skeptical of AI-generated content. Generic AI outputs tend to be structurally flat, logically sloppy, and formulaic in ways that experienced clients will notice immediately. A SWOT analysis that lists "strong brand" as a strength and "economic uncertainty" as a threat is not an analysis. It is a template populated with plausible-sounding words. A memo that buries the recommendation in the fourth paragraph and calls the structure "comprehensive" is not Minto. It is the kind of writing Minto was invented to replace.

What changes with framework-specific AI skills: each skill encodes a particular framework's logic as its operating constraint. The Minto skill does not write documents. It applies the Pyramid Principle. The PESTEL skill does not summarize news. It structures macro intelligence into the six specific forces and derives actionable implications. The strategy skill does not give generic advice. It applies ten named frameworks in sequence and produces outputs in the format each framework requires.

The result is output that reads like it was produced by someone who knows the frameworks, because the skill enforces them. The consultant's job is to bring the specific context (the client's situation, the market dynamics, the strategic questions) and direct the framework application. The skill handles the structure.

1. Minto Pyramid: Find the Logic Gaps Before Your Client Does

This skill applies Barbara Minto's Pyramid Principle to any draft: extracts the one-sentence governing thought, diagnoses MECE argument gaps, and delivers a color-coded visual restructuring plan that shows exactly where the logic breaks down and how to fix it.

The Pyramid Principle is the communication standard that McKinsey, BCG, and Bain train into every analyst. The core discipline: answer first, support with three to five mutually exclusive and collectively exhaustive arguments, evidence beneath each argument. Most documents violate this structure in the same ways: the answer is buried because the writer discovered it at the end and never went back to restructure; two arguments overlap because they were written on different days; a critical evidence gap exists that the writer did not notice because they were too close to the material.

The skill catches all three before the client does.

"Minto this memo. [paste draft] The memo argues we should enter the German market, but I want to confirm the logic holds before it goes to the steering committee. Flag any MECE violations in the supporting arguments, identify where the governing thought currently appears versus where it should appear, highlight any evidence gaps, and give me a replacement opener sentence that leads with the answer."

Before: The governing thought is buried in paragraph four. Two supporting arguments overlap. The steering committee notices in the review meeting and the presenter cannot defend the structure on the spot. A critical evidence gap surfaces only when a partner asks a question the memo never addressed.

After: Color-coded pyramid showing the answer buried in paragraph four, two overlapping arguments collapsed into one, a missing evidence gap flagged in red, and an exact replacement opener sentence. Before the steering committee sees it.

Setup: 5 minutes. Works on memos, slide decks, executive summaries, proposals, and any document where structured argument matters.

2. PESTEL Analysis: Macro Intelligence Across All Six Forces

This skill runs a systematic macro-environmental scan across Political, Economic, Social, Technological, Environmental, and Legal forces, producing specific, sourced intelligence for each cell, deriving the top opportunities and threats, and generating strategic recommendations grounded in the macro picture.

PESTEL's weakness in practice is that it tends to produce vague outputs: "the regulatory environment is complex" in the Legal cell, "technological disruption continues" in the Technology cell. These observations satisfy the template but provide no strategic insight. The macro forces that actually matter for a specific company entering a specific market are precise: the EU AI Act transparency requirement that affects a particular product feature, the exchange rate trend that changes the unit economics of European pricing, the remote freelancer growth curve that represents the core demand driver.

The skill enforces specificity by requiring a defined subject (a company, a product, a market entry) and producing intelligence that is particular to that subject in each of the six cells.

"Run PESTEL for our AI-powered invoice automation product launching in Germany and France in Q3. We are a US-based Series B fintech, 85 employees. Focus on factors that would materially affect our go-to-market or unit economics, not general observations. For each cell, identify the two or three specific forces most relevant to us and their strategic implication. Conclude with the top three opportunities and top three threats this analysis surfaces, and three strategic recommendations for the EU launch."

Before: PESTEL built in a two-hour workshop produces six cells of generic observations. The Legal cell says "GDPR compliance required." The Technology cell says "AI adoption accelerating." No strategic implications are derived.

After: Political: EU AI Act Article 13 transparency requirements apply to our automated decision output, requires explainability layer before Q3 launch. Economic: Euro/USD at 1.08 compresses our EUR pricing by 8% versus last year's model, repricing needed. Legal: German B2B invoice requirements under GoBD create a compliance integration opportunity competitors haven't addressed. Three specific strategic recommendations derived from the intersection of forces.

Setup: 30 minutes (longer than other skills because quality PESTEL requires research input). Works for market entry analysis, annual strategy decks, investor materials, and regulatory risk assessments.

3. Business Strategy Consultant: Ten Frameworks in One Engagement

This skill applies ten named strategic frameworks in a single session: SWOT analysis, go-to-market strategy, pricing strategy review, KPI dashboard design, growth lever identification, competitive analysis, 30/60/90-day planning, and more, producing structured outputs in the format each framework requires.

The problem: the same ten frameworks get applied to almost every engagement, but applying them thoroughly requires hours of preparation per framework. A SWOT that is actually MECE (where the strengths are specific capabilities tied to evidence, not brand-name adjectives) takes time to construct. A pricing strategy that considers value-based, competitive, and cost-plus approaches and recommends a specific structure with rationale takes longer. A KPI dashboard that selects leading and lagging indicators tied to the specific business model and strategic priorities is not a fifteen-minute exercise.

The skill compresses that preparation. Each framework is applied in the format that makes it useful, not as a bullet list that vaguely gestures at the framework, but as the structured output the framework was designed to produce.

"Run a full strategic analysis on this SaaS business. Context: B2B project management tool, $3.2M ARR, 140% net revenue retention, losing in competitive evaluations to Asana on enterprise features, strong in SMB but growth slowing. Apply SWOT (specific, evidence-based), identify the top three growth levers with effort/impact estimates, review current pricing structure and recommend adjustments, design a KPI dashboard for the leadership team, and build a 90-day action plan prioritized by strategic impact."

Before: A two-day strategy off-site produces a deck with a SWOT that lists "strong team" as a strength and "competitive market" as a threat. The go-to-market section recommends "focusing on the enterprise segment" without naming the specific product gaps that need to close first. The 90-day plan has no effort estimates.

After: SWOT with evidence-backed specifics. Three growth levers ranked by effort/impact (move upmarket via enterprise tier: high impact, 6-month timeline; double SMB expansion motion: medium impact, immediate; partner channel: high impact, 12-month build). Pricing recommendation with rationale. KPI dashboard with leading and lagging indicators. 90-day plan with sequenced priorities.

Setup: 10 minutes. Works for startup strategy sessions, SMB consulting engagements, and internal strategy teams.

4. Proposal Generator: Proposals That Reflect What the Prospect Actually Said

This skill produces tailored proposals by analyzing discovery call notes, emails, and prospect conversations, addressing each specific objection by name, positioning against the competitor they mentioned, including the integration they asked about, and structuring pricing to the budget they stated.

The generic proposal problem is universal in consulting: the prospect raises three objections in the discovery call, mentions a competitor twice, asks specifically about one integration, and expresses skepticism about timeline. The proposal that arrives a week later addresses none of it. It is a modified version of the last proposal, with the company name updated and the scope section adjusted. The prospect reads it and concludes that the firm did not listen, because the proposal is evidence they did not.

The skill inverts the process. It starts with the prospect's words (the transcript, the email thread, the notes from the discovery call) and builds the proposal around what was actually said.

"Generate a proposal based on my discovery call with Meridian Capital. Key points from the call: they have tried two previous strategy consultants who delivered decks but no implementation support, they want someone who stays through execution; their CFO is skeptical about AI tools and asked specifically about data security; they mentioned comparing us to Deloitte on price; the timeline is critical because their board review is in 90 days. Our differentiation: we embed with the team through implementation, not just deck delivery. Structure the proposal to address each of these directly, and include a 90-day milestone plan that maps to their board review date."

Before: Proposal sent is a modified version of the last engagement's proposal. The CFO's data security concern is not addressed. The previous-consultant objection appears nowhere. Deloitte is not named. The timeline section says "8 to 12 weeks" without referencing the board review date they mentioned. Meridian goes with Deloitte.

After: Executive summary leads with the implementation-through-execution differentiator positioned directly against their previous consultant experience. CFO data security concern addressed in a dedicated section with specifics. Deloitte price comparison reframed on value basis. 90-day milestone plan anchored to the board review date they mentioned. Meridian feels heard.

Setup: 10 minutes. Works best when you paste actual discovery call notes or email threads. The more specific the prospect's words, the more precisely the proposal reflects them.

The Consulting Workflow: From Macro Scan to Signed Engagement

The four skills map onto a consulting engagement workflow end to end. PESTEL scans the macro environment and surfaces the strategic context. The strategy consultant applies the analytical frameworks that turn that context into a structured strategic picture. The Minto Pyramid ensures every deliverable (the situation memo, the strategy deck, the executive summary) leads with the answer and holds up to logical scrutiny. The proposal generator turns the discovery conversation into a proposal that wins the next engagement.

The common thread is framework fidelity. Consulting clients pay a premium for structured thinking, and structured thinking is only valuable when the structure is rigorous. A SWOT with generic entries, a memo that buries the recommendation, a proposal that ignores what the prospect said: these are failures of structure, not failures of intelligence. The skills enforce the structure so the intelligence is what the client sees.

Consultants who use these skills report the same effect: they spend less time building the framework and more time filling it with insight. The PESTEL cells get populated faster, so more time goes to sourcing the specific intelligence that makes each cell useful. The Minto diagnosis catches structural problems in draft, so the revision is a targeted fix rather than a full restructure at 11pm before the client meeting. The proposal reflects the discovery conversation, so the first draft requires fewer rounds of "make it more specific to them." The analytical work (the judgment about what the macro forces mean, which strategic levers are real, what the client's actual problem is) stays with the consultant. The scaffolding gets handled.

Getting Started

I publish all four consulting skills as free, downloadable templates at claudecodehq.com: Minto Pyramid logic diagnosis, PESTEL macro analysis, a 10-framework business strategy toolkit, and discovery-call-driven proposal generation. Each one is a single file you install once and use whenever you need it. Start with whichever framework you apply most often under the most time pressure. For most consultants, that's either Minto (if you write memos and decks weekly) or the proposal generator (if you're in business development mode).

Originally published on claudecodehq.com

I Run People Ops for 80 People. These 4 AI Skills Handle the Paperwork So I Can Do the People Work.

Daniel Marin — Thu, 18 Jun 2026 18:44:44 +0000

Structured onboarding programs, balanced performance reviews, compliant handbook updates, and automated HR workflows. How to stop drowning in documentation and start spending time on the conversations that matter.

HR managers are asked to do two fundamentally different things simultaneously: build a culture of trust and human connection, and maintain an enormous volume of documentation that underpins every people process in the organization. These two demands are in constant tension. The documentation work (onboarding checklists, performance reviews, policy updates) consumes exactly the time and cognitive bandwidth that should be going toward the human work.

The documentation problem compounds in a specific way. Onboarding programs drift because the Google Doc nobody owns gets out of date. Performance reviews go generic because managers write eight blank evaluations the week they are due and default to vague praise. The employee handbook still references the 2022 PTO policy and the office that closed last year because handbook updates feel urgent only the week a new hire asks why the policy they received doesn't match what they were told in the interview. And the underlying workflows (account provisioning, buddy assignment, orientation scheduling) run on tribal knowledge rather than repeatable process.

AI skills handle the documentation layer. HR managers keep ownership of the human decisions (who to hire, how to develop someone, what culture to build) while the paperwork gets handled.

The Documentation Tax on People Work

Every HR process that matters to employees (joining the company, getting feedback on their performance, understanding their rights and benefits) is mediated by documentation. When that documentation is good, employees feel supported and the company runs smoothly. When it is bad, the damage compounds: a disorganized first week drives early attrition, a vague performance review fails to develop the employee and creates legal exposure if the relationship later deteriorates, an outdated handbook creates compliance risk the company does not know it is carrying.

The paradox is that producing good HR documentation requires sustained attention at exactly the moments when HR managers are busiest. Onboarding documentation needs to be prepared before a new hire's first day, which is also when the offer was just accepted and multiple stakeholders are asking for status updates. Performance review season requires thoughtful, specific feedback for every direct report, which is also when the quarter is closing and headcount planning is underway.

The skills below compress the documentation work so the attention it requires does not crowd out the relationship work it is supposed to support.

1. Onboarding Documentation: A First Week That Actually Works

This skill builds a complete onboarding program: pre-start checklist covering IT, access, and equipment; a day-by-day schedule for weeks one through four; a role-specific technical setup guide; a 30/60/90-day milestone framework; buddy program guidelines; and manager check-in templates.

The failure mode it addresses: a new hire's first week is a mess of forgotten laptop requests, missing access permissions, and "ask Sarah, she knows how that works." Bad onboarding doubles ramp time and drives early attrition. The cost of losing someone in the first ninety days (recruiting fees, manager time, lost productivity, the impact on team morale) easily exceeds six months of that employee's salary. The root cause is almost always a lack of structure, not a lack of goodwill.

"Build an onboarding program for a new Senior Product Manager joining our 80-person SaaS company on July 7th. She will report to the VP Product, own the growth product line, and work closely with engineering leads across two teams. We use Notion, Linear, Figma, and Slack. Pre-start checklist for IT and access, day-by-day schedule for weeks 1 to 4, a 30/60/90 milestone framework with specific success criteria, buddy program guidelines, and manager check-in templates for weeks 1, 4, and 12."

Before: New hire shows up on day one, laptop isn't ready, Notion access is pending, nobody told engineering she was starting. The first week is improvised. By day 30 she has not met two of her key stakeholders. Early attrition risk is already elevated.

After: Pre-start checklist sent two weeks before day one. Day-by-day schedule through week four. 30/60/90 milestones agreed with the manager before the start date. Buddy assigned. Check-in templates ready. The first week runs off a plan, not tribal knowledge.

Setup: 5 minutes. Works for any role and team size. Reuse the same program structure across hires in the same role.

2. Performance Review Writer: Specific Feedback Without the Blank-Page Paralysis

This skill generates thorough, balanced performance reviews with specific accomplishments and metrics, rating justifications aligned to rubric, strength areas with examples, development areas with actionable suggestions, and a 90-day growth plan.

Review season creates a specific failure mode: managers write eight blank reviews the week they are due. The reviews that emerge are either generically positive ("great team player, strong communicator") or vague about development areas in ways that provide no actual guidance. Both outcomes fail the employee. The generic praise does not help them understand what to keep doing, and the vague development feedback does not tell them what to change.

The skill requires the manager to provide specific input: the accomplishments they observed, the behaviors they want to reinforce, the gaps they have seen. It then structures that input into a review that meets the formal requirements of a performance document: specific, measurable language; examples tied to outcomes; rating justification that would hold up in a calibration conversation; development recommendations with clear next actions.

"Write a performance review for Marcus, a senior software engineer. Strengths: delivered the payments API refactor three weeks ahead of schedule with zero production incidents, consistently unblocks junior engineers without being asked, excellent written technical documentation. Development area: communication in cross-functional meetings, tends to go quiet when non-engineers are present, misses opportunities to advocate for technical constraints early. Rating: Exceeds expectations on delivery, meets expectations on collaboration. Include a 90-day growth plan focused on the communication gap."

Before: Manager writes eight reviews in two days, all starting to blur together. Marcus gets "strong technical skills, great team player" with a vague note about "improving stakeholder communication." He leaves the review unsure what to actually change or how.

After: Balanced review with specific accomplishments tied to outcomes, rating justification for both dimensions, strength areas with examples, communication development area with concrete next actions, and a 90-day growth plan Marcus can act on from day one.

Setup: 10 minutes. Works for annual reviews, quarterly check-ins, and mid-cycle feedback.

3. Employee Handbook Updater: Compliance-Aligned Policies in Plain Language

This skill drafts and updates handbook sections and HR policies with consistent formatting, plain language, compliance notes for relevant state laws, and a change log documenting what was updated and why.

Handbook neglect has a predictable pattern: a company launches with a handbook assembled from templates, updates the PTO policy in 2023, adds a remote work section in 2024, but never updates the rest of the document to reflect the changes. The handbook now contains internally contradictory information. The dress code section still describes the pre-pandemic office. The section on office hours predates the shift to flexible scheduling.

The legal exposure is the less visible problem. Employment law changes at the state level faster than most HR teams can track (leave requirements, pay transparency obligations, non-compete enforceability) and an outdated handbook can inadvertently make promises the company does not keep or fail to make disclosures the company is legally required to provide.

"Update our employee handbook with three policy changes: (1) PTO policy changed from accrual-based to unlimited, with a minimum of 10 days encouraged; (2) remote work policy updated to allow fully remote for all roles except those requiring physical presence; (3) parental leave expanded from 8 weeks to 16 weeks for primary caregivers and 4 weeks for secondary caregivers. We operate in California, New York, and Texas. Match the formatting of the existing sections I will paste below. Include a change log and flag any state-law considerations for each policy."

Before: Three policy changes approved, three Slack messages sent, nobody updated the handbook. Six months later a new hire asks a question, HR checks the handbook, finds it still says the old policy. A California employee files an inquiry because the unlimited PTO policy wasn't documented correctly under CA law.

After: Three updated handbook sections in consistent formatting, plain-language policy explanations, California and New York compliance flags for the PTO and parental leave changes, and a change log documenting version, date, and nature of each update. Ready for legal review before publishing.

Setup: 10 minutes. The compliance flag output is not legal advice. It identifies areas to review with counsel, not a substitute for it.

4. HR Process Automation: Replace Tribal Knowledge With Repeatable Workflows

This skill designs and implements full HR workflow automation: new hire account creation across all tools, welcome email sequences, onboarding buddy assignment, orientation meeting scheduling, offboarding access revocation, and pipeline tracking. Replacing the 47-step Google Doc with a workflow that runs automatically.

This skill addresses the layer beneath the documentation: the manual work that executes the documentation. An onboarding checklist is only as good as the process that ensures each item gets done. If account provisioning requires a Slack message to IT, a separate request to the engineering manager, and a reminder email three days later because nobody saw the first message, the onboarding program is still running on tribal knowledge. Just tribal knowledge with a nicer document attached.

"Design an automated employee onboarding workflow for our company. Tools we use: Google Workspace for email and calendar, Notion for documentation, GitHub for engineering, Figma for design, Slack for communication, Rippling for HRIS. When a new hire record is created in Rippling: create Google Workspace account, add to relevant Slack channels based on department, provision Notion access, add to GitHub org if engineering hire, send welcome email with first-day logistics, assign onboarding buddy from the same team, and schedule a kickoff meeting with their manager for day one. Track completion status for each step."

Before: IT provisioning is a Slack thread. The buddy assignment is a calendar invite that sometimes gets missed. The welcome email is sent manually. Engineering hires miss GitHub access half the time until they ask. Every onboarding is slightly different based on who handled it.

After: New hire record created in Rippling triggers the full workflow automatically: accounts provisioned across all tools, Slack channels joined, welcome email sent, buddy assigned, day-one kickoff scheduled, completion tracked. Every onboarding is identical.

Setup: 10 minutes to design the workflow. Implementation time varies by tool stack and automation platform. This skill is more technical than the other three.

The HR System: Documentation That Stays Current

The four skills form a complete people-ops documentation system. The onboarding skill defines what a great first week looks like for each role. The process automation skill runs that program without manual coordination at every step. The performance review writer ensures that the feedback employees receive at review time is specific and actionable rather than generic. The handbook updater keeps the policy foundation current with every change, in a format that employees can read and that holds up to compliance scrutiny.

What the system replaces is not HR judgment. It is HR overhead. The decision about what a good onboarding program should include, what feedback is honest and constructive, what policies are appropriate for the company's culture and legal obligations: those stay with the HR manager. What gets automated is the blank-page problem. The paralysis of building from scratch under time pressure, the inconsistency of tribal knowledge, the documentation drift that makes every HR document gradually less accurate over time.

HR managers who use these skills report the same shift: they spend less time staring at blank documents and more time in the conversations that actually require their expertise. The new hire who is struggling in week three. The manager who needs coaching on how to deliver difficult feedback. The policy question that requires judgment rather than documentation. The documentation layer runs. The people layer gets the attention it deserves.

Getting Started

I publish all four HR skills as free, downloadable templates at claudecodehq.com: onboarding documentation and checklists, performance review writing, employee handbook updates with compliance flags, and HR process automation design. Each one is a single file you install once and use whenever you need it. Start with whichever process is causing the most pain this quarter. For most HR teams, that's either onboarding (if you have hires starting soon) or performance reviews (if review season is approaching).

Originally published on claudecodehq.com

I Screen 200 Resumes in One Session Now. Here Are the 4 AI Skills That Changed How I Recruit.

Daniel Marin — Wed, 17 Jun 2026 18:41:02 +0000

Ranked shortlists with match scores, job descriptions that actually attract top candidates, pipeline tracking with stalled-candidate alerts, and comp benchmarks that replace gut-feel counteroffers with market data.

Recruiting is one of the highest-leverage jobs in any organization. Hire the right people and everything else gets easier. Hire wrong and problems compound for years. It is also one of the most operationally exhausting: 200 resumes for one role, 15 open requisitions simultaneously, hiring managers asking for pipeline updates daily, and a compensation decision that needs to be made in 24 hours against a competing offer.

The structural problem is that the work most likely to lose a great candidate (slow follow-up, a generic job description that doesn't stand out, a screening process that lets good-fit applicants fall through) is the work that feels least urgent when there are 200 resumes in the inbox. Recruiters end up spending their highest-attention hours on screening volume, leaving the strategic work (writing a compelling JD, building a real pipeline view, making a competitive offer) for the end of the day when the bandwidth is gone.

AI skills invert that. The four skills below handle the high-volume mechanical layer of recruiting: screening 200 resumes in one session, generating a JD that actually differentiates the role, tracking pipeline health across all open roles, and benchmarking comp against real market data. The recruiter's attention goes to the work that requires human judgment: relationships, culture reads, negotiation.

Why Recruiting Gets Stuck in the Volume Trap

The recruiting volume trap works like this: a role opens, applications arrive, the recruiter starts screening manually. By resume fifty, standards are drifting. Not intentionally, but because humans cannot maintain consistent evaluation criteria across 200 documents. By resume one hundred, the recruiter is pattern-matching on surface signals (school name, company brand) rather than actual job-fit indicators. The shortlist that emerges is not the most qualified fifteen candidates. It is the fifteen who happened to appear when the recruiter was paying attention.

Meanwhile, the job description that drove those 200 applications was written in two hours from a template. It reads like every other posting on LinkedIn. The candidates who would have been the best fit saw nothing distinctive about the role and applied to three other companies that week.

The pipeline tracker is a spreadsheet with color-coded cells that nobody updates consistently, so the hiring manager's weekly "where are we on this?" question cannot be answered without fifteen minutes of manual reconciliation.

And when the finalist candidate comes back with a competing offer, the recruiter makes a counteroffer based on what the budget allows plus a rough sense of market rates. Not a benchmarked analysis of where the candidate's comp would land against the 50th percentile for that role in that market.

Every one of these failure modes is a mechanical problem, not a judgment problem.

1. Applicant Screening: Ranked Shortlist From 200 Resumes

This skill scores job applications against your specific requirements, assigns match percentages, flags red flags, and produces a ranked shortlist of your top candidates with interview talking points for each.

The key difference from manual screening: the criteria stay consistent from candidate 1 to candidate 200. The skill evaluates every application against the same requirements in the same order with the same weighting. No fatigue, no pattern-matching drift, no unconscious preference for familiar brand names. A strong candidate who applied on day three does not get screened differently than an identical candidate who applied on day one.

The output is not a simple pass/fail. It is a scored shortlist with the reasoning visible: which requirements each candidate meets, which they fall short on, and what an interviewer should probe in a first-round conversation.

"Screen these 150 applications for our Senior React Developer role. Must-haves: 5+ years React, TypeScript, experience with a production codebase over 100k LOC. Nice-to-haves: GraphQL, AWS, prior startup experience. Hard disqualifiers: no professional React experience, only bootcamp projects. Score each candidate 0 to 100 against these criteria, flag any red flags in the application, and give me a ranked top 15 with one-line interview talking points per candidate."

Before: Three days reading 150 resumes manually. By resume 60 the standards are drifting. The shortlist reflects who caught the recruiter's eye, not who best fits the role. Two strong candidates applied on day three and never got a close read.

After: Every candidate scored consistently against the same criteria. Ranked top 15 with match percentages, red flags flagged, hard disqualifiers automatically removed. Interview talking points ready for the hiring manager briefing. In one session.

Setup: 10 minutes. Works for any role and volume.

2. Job Description Generator: JDs That Actually Attract Top Candidates

This skill produces a compelling job description, structured interview rubric with scoring criteria, hiring timeline, and evaluation scorecard. A complete hiring package designed to attract strong candidates, not just collect applications.

The problem with most job descriptions is not that they contain wrong information. It is that they contain the same information as every other posting for the role. A Senior Product Manager JD that lists "cross-functional collaboration," "data-driven decision making," and "5+ years of experience" looks identical to 200 other Senior PM postings. The candidates who would be a great fit see nothing to distinguish it.

The skill asks for context that forces differentiation: what makes this role different from the equivalent role at a competitor, what the first 90 days actually look like, what the growth trajectory is, what kind of person has thrived in this team. The output uses that context to write a JD that communicates the genuine opportunity, not just the requirements.

"Create a complete hiring package for a Senior Product Manager, Payments. Context: we are a Series B fintech, 120 people, the payments product is our core revenue driver and this PM will own the full roadmap. The previous PM left to join a later-stage company. We need someone who wants to build, not manage. The growth path is VP Product within 18 months if the roadmap executes. Include a compelling JD, a structured interview rubric with scoring criteria, and a 6-week hiring timeline."

Before: Generic JD assembled from a template in two hours. Posted to LinkedIn, gets 200 applications, 160 of which don't meet basic requirements. The three candidates who would have been perfect fits saw nothing to make the role stand out and didn't apply.

After: Differentiated JD that communicates the genuine opportunity, VP growth path, and what success looks like in 90 days. Interview rubric with scoring criteria. 6-week hiring timeline. The complete hiring package ready to share with the hiring manager before posting.

Setup: 5 minutes. Works for any role, level, or function.

3. Pipeline Tracker: Visibility Across All Open Roles

This skill tracks candidate status across all open roles with stage-by-stage conversion rates, time-in-stage alerts for stalled candidates, hiring manager dashboards, weekly pipeline health reports, and sourcing channel effectiveness analysis.

The failure mode it addresses: your best candidates ghost you because nobody followed up for five days. Not because the recruiter forgot. Because with 15 open roles, a candidate who moved to the offer stage on role twelve is not at the top of anyone's mental model when they are spending Tuesday reviewing applications for role seven. The pipeline tracker surfaces who is stalled, in which stage, for how long.

The sourcing channel analysis is equally valuable. If LinkedIn is generating 80% of applications but only 20% of hires, and employee referrals are generating 15% of applications but 40% of hires, that ratio should be driving sourcing investment decisions. But only if someone has done the analysis. The skill calculates it automatically from pipeline data.

"Set up a recruiting pipeline tracker for our 12 open roles. We track five stages: applied, phone screen, hiring manager interview, panel, offer. I need: stage-by-stage conversion rates per role, a flag for any candidate who has been in the same stage for more than 5 business days, a hiring manager summary showing role status and next actions, and a sourcing channel analysis breaking down where our hires are actually coming from versus where our applications are coming from."

Before: Pipeline lives in a spreadsheet nobody updates consistently. Hiring manager asks "where are we on the backend role?" and the answer requires 15 minutes of manual reconciliation. A finalist candidate went dark because the follow-up fell through a gap between recruiter and coordinator.

After: Stage conversion rates per role, stalled-candidate alerts at 5 days, hiring manager dashboard with role status and next actions, weekly pipeline health report, and sourcing channel effectiveness analysis. All from a single pipeline update session.

Setup: 10 minutes. Works across any number of open roles and any pipeline stage structure.

4. Compensation Benchmarker: Market Data Before the Counteroffer Conversation

This skill produces percentile-based compensation benchmarks (25th to 90th) for any role, band placement analysis, equity grant modeling, retention risk flags for underpaid employees, and market adjustment recommendations.

The gut-feel comp problem: a finalist candidate comes back with a competing offer. The recruiter checks the internal budget, knows roughly what the market pays from memory, and makes a counteroffer. It lands somewhere defensible. But "defensible" is not the same as competitive. If the offer is below the 50th percentile for the role in that market, the candidate takes the competing offer and tells their network that your company underpays.

The skill replaces the gut-feel decision with a structured analysis: where does the candidate's current comp and the competing offer land against the 25th, 50th, 75th, and 90th percentile for that role, level, and location. What equity grant would be required to make the total comp competitive at the 65th percentile. Which existing employees in the same band are below the 40th percentile and represent a retention risk that should be addressed before the next annual review.

"Benchmark compensation for a Senior Software Engineer, IC4 level, in San Francisco. The candidate is currently at $195k base, $50k equity per year. The competing offer is $230k base, $80k equity, $30k signing. Our internal band for this level is $180k to $220k. I need: where the candidate and the competing offer land against SF market percentiles, what total comp at our 75th percentile looks like, whether a signing bonus can bridge the gap without adjusting base, and retention risk flags for our current IC4s if this band hasn't been refreshed since 2023."

Before: Counteroffer built from budget ceiling plus Glassdoor estimates. The candidate takes the competing offer. Three months later, another senior engineer gets a competing offer in the same band and the same conversation repeats because the band still hasn't been refreshed.

After: Competing offer placed at the 78th percentile SF market. Counter-strategy: $215k base + $25k signing lands at 70th percentile total comp. IC4 band flagged as below 40th percentile for three existing employees. Retention risk memo ready for the next comp review cycle.

Setup: 10 minutes. Works for individual offer decisions and full-team comp reviews.

The Recruiting System: From Requisition Open to Offer Accepted

The four skills form a complete recruiting workflow. A role opens: the job description generator produces a differentiated JD with an interview rubric in one session. Applications arrive: the screening skill scores the full pool and produces a ranked shortlist in hours, not days. Candidates move through stages: the pipeline tracker surfaces stalled candidates before they go dark and gives the hiring manager real-time visibility without a status meeting. A finalist emerges: the comp benchmarker produces the offer strategy before the counteroffer conversation happens, not during it.

The consistent thread across all four skills is that the mechanical work (reading 200 resumes against the same criteria, maintaining a pipeline spreadsheet, Googling salary data) gets handled without consuming recruiter attention. The work that actually requires a recruiter (reading candidates as people, building relationships, reading whether a candidate is genuinely excited or just shopping offers) gets more time and better attention because the operational baseline is already covered.

Getting Started

I publish all four recruiting skills as free, downloadable templates at claudecodehq.com: applicant screening with match scoring, differentiated job description generation, pipeline tracking with stalled-candidate alerts, and percentile-based compensation benchmarking. Each one is a single file you install once and use whenever you need it. Start with the screening skill the next time a role opens and 200 resumes land in your inbox. The consistency alone is worth the setup.

Originally published on claudecodehq.com

I Cover 15 Stocks. These 4 AI Skills Cut My Earnings Update From 3 Hours to 45 Minutes.

Daniel Marin — Tue, 16 Jun 2026 17:27:29 +0000

Post-earnings updates, morning notes, thesis tracking, and initiation reports. How equity research analysts compress the structural work so the analysis gets the time it deserves.

Equity research is a deadline business dressed up as an intellectual one. Earnings drop after the close. The morning meeting is at 7am. The initiation report that took six weeks of modeling and writing lands the same week as two other companies in your coverage universe reporting. The work is genuinely analytical, but the container it lives in is brutal.

The part of equity research that AI can help with is not the part analysts worry about. No serious analyst is worried that AI will replace their read on management credibility, their sense of whether a new product cycle is real, or their view on valuation in a sector they have covered for a decade. What actually consumes analyst time is structural: building the earnings update framework from scratch at 5pm, drafting the morning note at 6am, maintaining thesis scorecards across a 15-stock coverage universe, and assembling the 40-page initiation report from a dozen separate working documents.

AI skills handle the structural layer. The analysis stays with the analyst. The scaffolding gets handled.

Why the Research Workflow Is Structurally Broken

Sell-side and buy-side analysts face the same underlying problem: the outputs of equity research are highly standardized (earnings updates follow the same structure at every firm, morning notes hit the same elements every day, initiation reports follow the same architecture across all banks), but the process for producing them is almost entirely unstructured. Every analyst rebuilds the template from memory, every time, under time pressure.

The earnings update is the sharpest example. A company reports after the close. The analyst has 12 to 18 hours to publish a credible update: revenue beat/miss, EPS beat/miss, guidance versus consensus, updated model, revised price target, and a thesis read. Miss something in that structure (skip the guidance commentary, omit the segment breakdown, bury the thesis implication) and the note reads as incomplete. Build the framework correctly under time pressure and the actual analysis takes an hour. Build the framework slowly and the analysis gets rushed.

The four skills below give analysts the frameworks pre-built and professional. The analyst brings the numbers, the read on management commentary, and the thesis judgment. The skill provides the structure, the output format, and the institutional conventions.

1. Earnings Update Report: 24-Hour Deadline, Professional Output

This skill produces a professional 8 to 12 page earnings update with beat/miss analysis, updated estimates table (old versus new), revised price target, and thesis read, structured for immediate publication.

The skill handles what takes the most time under deadline pressure: the framework. Revenue beat/miss versus consensus and your estimate. Segment-level breakdown. EPS bridge. Guidance versus street. Updated model with old-versus-new estimates table. Revised price target with methodology. Thesis confirmation or thesis change. Twelve charts. Each section has a professional structure the skill enforces, so the analyst is filling in analysis, not building architecture at 9pm.

"Create an earnings update for Nike Q2 FY24. Revenue was $12.4B vs. $12.6B consensus, a miss. EPS was $1.03 vs. $0.85 consensus, a beat driven by gross margin expansion to 44.6%. Management guided Q3 revenue flat to down 1% versus street at +2%. Inventory is down 14% year-over-year. My thesis is that the DTC mix shift is working but the wholesale channel remains a headwind into FY25. Update my price target from $110 to $105. Format as a full earnings update with an old/new estimates table."

Before: Two hours building the earnings update structure from a blank document after a late-night earnings call. The framework eats the time the analysis needed. The note goes out at midnight, rough around the edges.

After: Professional 8 to 12 page earnings update with beat/miss analysis, old/new estimates table, 8 to 12 charts, revised price target, and thesis read. Framework complete in minutes. Analyst fills in the actual numbers and insight.

Setup: 5 minutes. Works for sell-side earnings updates and buy-side post-earnings position reviews.

2. Morning Note: Actionable Intelligence Before the Open

This skill drafts the pre-market morning note: overnight developments on coverage stocks, earnings reactions, key events for the trading day, and a long/short trade idea with risk parameters.

The morning note is not a news summary. PMs and traders already read the news. The value of a morning note is the analyst's opinionated take: what overnight data means for your specific thesis on your specific stocks, and a trade idea with a clear entry rationale and risk frame. The skill enforces that discipline: top call headline, overnight developments with thesis implications, events today with positioning read, long/short idea with entry level and stop.

At 6am, with a 7am meeting hard stop, the morning note cannot be built from scratch. The skill provides the structure instantly so the analyst can focus on the overnight data that actually matters for the coverage universe.

"Draft my morning note for tech coverage. Overnight: NVDA reported above consensus on data center, guiding up 10% for next quarter, stock up 6% in after-hours. MSFT Azure growth reaccelerated to 31% versus 28% last quarter. Today: Fed speakers at 10am and 2pm, CPI tomorrow. My call: the AI capex cycle is broadening from hyperscaler to enterprise. SMCI is the asymmetric long into CPI. Draft with a SMCI long idea, stop at the 200-day."

Before: Writing the morning note by hand at 6am from a blank page. The structure is fine by muscle memory, but the trade idea section always gets rushed. The note is tight on some days and generic on others depending on how tired the analyst is.

After: Top call headline, overnight developments with thesis implications, key events with positioning read, and a SMCI long idea with entry rationale, risk parameters, and stop level. One-page note ready before the morning meeting.

Setup: 5 minutes. Works for sell-side morning notes distributed to sales desks and buy-side pre-meeting briefings.

3. Thesis Tracker: Conviction Monitoring Before It Becomes a Portfolio Mistake

This skill maintains structured investment theses with scorecard tracking across thesis pillars, a catalyst calendar, and conviction-level monitoring, updated with each new data point.

Thesis drift is how positions outlast their rationale. An analyst initiates on a company with a three-pillar thesis: margin expansion driven by mix shift, a new product cycle in 18 months, and international expansion. Twelve months later, mix shift is working, the product cycle slipped to 24 months, and international is a question mark. The position is still in the book. The original initiation recommendation still shows on the coverage page. The thesis has not been formally reassessed.

The skill forces the reassessment. Each pillar of the original thesis has a status (on track, behind, ahead, broken) and a confidence score. New data points are logged with their thesis impact assessed immediately, not accumulated in a mental model that degrades over time. The catalyst calendar tracks what would confirm or refute each pillar. When conviction shifts, the skill captures when and why.

"Update my investment thesis for Apple. Original thesis: services margin expansion to 75%+ by FY26, India manufacturing as a China supply chain hedge, and Vision Pro as an option on spatial computing. Update with latest earnings: services gross margin hit 73.9%, India manufacturing now at 14% of iPhone production (ahead of plan), Vision Pro sales are tracking below break-even unit economics. Revise my conviction scorecard and flag whether Vision Pro has become a thesis impairment or remains an option position."

Before: Thesis lives in the analyst's head and a six-month-old initiation report. New data points accumulate without formal thesis impact assessment. Conviction drifts without anyone noticing. Position is still held six months after the original thesis broke on one pillar.

After: Updated thesis scorecard showing each pillar status, Vision Pro flagged as an option position not a thesis impairment with explicit reasoning, revised conviction level, and a catalyst calendar showing what data would change the call in either direction.

Setup: 5 minutes. Works across a full coverage universe. Run it after every earnings report, every major data point, and every management meeting.

4. Initiating Coverage: Institutional-Quality Reports Without the Six-Week Grind

This skill builds a 30 to 50 page institutional initiation report through a structured 5-task workflow: company research, financial modeling, valuation, chart creation, and report assembly, producing 10,000 to 15,000 words, 25 to 35 charts, 12 to 20 tables, DCF valuation, comps analysis, and a BUY/HOLD/SELL recommendation.

An initiation report is the most labor-intensive output in equity research. It requires synthesizing company history, industry dynamics, competitive positioning, financial model, valuation, and a forward thesis into a coherent narrative that reads as authoritative to both portfolio managers and company management. The structural work (the framework of sections, the chart architecture, the comps table format, the DCF model structure) is the same every time. The intellectual work (the thesis, the variant view, the valuation argument) is unique to each company.

The 5-task workflow separates these cleanly. Task 1 builds the company and industry research foundation. Task 2 runs the financial model. Task 3 builds valuation (DCF plus comps). Task 4 creates the chart package. Task 5 assembles the full report. Each task produces institutional-quality intermediate output. The analyst's judgment goes into the thesis, the variant view, and the valuation call, not into formatting tables and building chart templates.

"Initiate coverage on Palantir Technologies (PLTR) with a BUY rating. My thesis: PLTR is the only pure-play enterprise AI platform with a proven government track record, and the AIP commercial inflection is underappreciated by the market. Key variant view: consensus models SBC normalization incorrectly, adjusted FCF is 40% above street estimates. Target price $35. Run the full 5-task workflow: company research, model, valuation (DCF + SaaS comps), charts, and report assembly."

Before: Six weeks for a senior analyst and one associate to produce a 40-page initiation. Four weeks of that is structural: building the model, formatting the comps, creating the chart package, assembling the sections. Two weeks is actual thesis development.

After: 30 to 50 page institutional-quality initiation with 10,000 to 15,000 words, 25 to 35 charts, DCF and comps valuation, and a BUY recommendation built through a structured 5-task workflow. The analyst's time goes to the thesis, not the template.

Setup: 10 minutes. Works for sell-side initiation reports and buy-side deep-dive research memos. The 5-task structure produces publishable intermediate outputs at each stage.

The Research System: A Coverage Universe That Runs Like a Practice

The four skills form a research system. Initiation coverage builds the original thesis and scorecard. The thesis tracker maintains it. The earnings update revises the model and price target after every quarterly print. The morning note communicates the current conviction read to the desk every morning with an actionable trade idea.

An analyst running a 15-stock coverage universe with this system operates differently from one who does not. The initiation creates a structured thesis document with explicit pillars and a catalyst calendar. After earnings, the update revises the model and price target in under an hour. The thesis tracker logs the earnings data point and assesses thesis impact immediately, not three weeks later when memory has degraded. The morning note communicates a genuine read on overnight data rather than a generic summary.

The result is a coverage practice where every position has a current thesis document, where conviction changes are logged and dated, where the morning note reflects a real analytical read, and where an initiation can be published in days rather than weeks. The intellectual work (the analysis, the judgment, the variant view) stays with the analyst. The structural work gets handled.

Getting Started

I publish all four equity research skills as free, downloadable templates at claudecodehq.com: post-earnings updates, pre-market morning notes, thesis tracking with conviction monitoring, and the full 5-task initiation report workflow. Each one is a single file you install once and use whenever you need it. Start with the earnings update before the next company in your coverage reports, and see how much time the framework saves when you're not building it from scratch at 9pm.

Originally published on claudecodehq.com

Our Security Team Was Drowning in Framework Documentation. These 4 AI Skills Fixed It.

Daniel Marin — Mon, 15 Jun 2026 19:55:15 +0000

CIS Controls gap assessments, NIST CSF profile building, PCI DSS CDE scoping, and FedRAMP SSP documentation. How to map controls and pass audits faster.

Security teams are drowning in framework documentation. NIST CSF 2.0, CIS Controls v8, PCI DSS v4.0.1, FedRAMP: each framework runs hundreds of pages of controls, sub-requirements, and implementation guidance, all using different vocabulary for overlapping concepts. A security engineer who needs to map their current posture against two frameworks simultaneously faces weeks of cross-referencing before they can write a single gap report.

The problem is not technical complexity. Experienced security teams know what access control means in their environment. What kills their time is the documentation work: scoping Implementation Groups, building current-versus-target profiles, determining which SAQ applies to their cardholder data environment, producing a Statement of Applicability. This is reading and writing work, not security engineering work.

AI skills handle the reading and writing. Security engineers stay focused on the actual risk decisions. The framework navigation and documentation get handled.

The Framework Proliferation Problem

Every security team faces the same structural tension: boards and enterprise customers demand proof of a structured security program, but the frameworks that provide that proof were written for regulators, not practitioners. CIS Controls v8 has 18 control families and 153 individual safeguards. NIST CSF 2.0 spans six functions and dozens of subcategories. PCI DSS v4.0.1 has 12 requirements with hundreds of testing procedures. FedRAMP Moderate references 325 NIST SP 800-53 controls.

These frameworks overlap significantly (access control, logging, incident response, and vendor risk appear in all of them) but each uses different clause numbering, different terminology, and different evidence requirements. A team pursuing both SOC 2 and ISO 27001 already maintains two parallel control inventories. Add CIS Controls and PCI DSS and the documentation overhead becomes a part-time job.

The skills below do not replace security judgment. They do replace the work of knowing which Implementation Group to scope, how to phrase a current profile, which SAQ type applies to a redirect-only integration, and how to structure an SSP outline. That is the work that eats security engineers' weeks without making the organization any more secure.

1. CIS Controls v8 Advisor: Scoped Gap Assessments With Framework Crosswalks

This skill scopes your Implementation Group (IG1, IG2, or IG3), runs a safeguard-level gap assessment across the CIS Top 18, and maps your controls to NIST CSF, ISO 27001, SOC 2, and CMMC.

The scoping step matters more than most teams realize. IG1 covers basic cyber hygiene for organizations of any size: 56 safeguards. IG2 adds controls for organizations managing sensitive data: 74 more. IG3 is the full set for mature programs: 23 additional. Starting with IG3 when IG1 is the right scope wastes months of remediation effort on controls that do not apply to your risk profile.

The crosswalk output is equally valuable: if you are already pursuing SOC 2 or ISO 27001, knowing which CIS safeguards you already satisfy through existing controls eliminates redundant work and lets you present a unified security narrative to auditors and enterprise customers.

"Scope our CIS Controls v8 Implementation Group. We are a 75-person SaaS company handling PII but not payment cards. We have MFA, endpoint protection, and a vulnerability scanner but no formal asset inventory or logging program. Run a gap assessment against the appropriate IG, prioritize the top 10 remediation items by risk impact, and map our existing controls to their NIST CSF equivalents."

Before: Three weeks manually reading the CIS Controls v8 guide, cross-referencing 153 safeguards against current controls, arguing internally about which IG applies, and producing a gap spreadsheet that is already outdated by the time it is finished.

After: IG2 scoping confirmed, safeguard-level gap report across all 18 controls, top 10 remediation priorities ranked by risk impact, and a crosswalk showing which existing controls satisfy NIST CSF subcategories. Produced in one session.

Setup: 10 minutes. Works for organizations at any maturity level.

2. NIST CSF 2.0 Advisor: Current-to-Target Profile and Board-Ready Risk Reporting

This skill assesses cybersecurity posture against the NIST CSF 2.0 six functions (Govern, Identify, Protect, Detect, Respond, Recover), builds current and target profiles, assigns implementation tiers, and maps the framework to other standards.

NIST CSF is the common language boards and regulators use to discuss cyber risk. CSF 2.0 added Govern as a sixth function, making cybersecurity governance an explicit first-class concern. If a CISO cannot show a current profile and a target profile with a clear gap narrative, they cannot credibly demonstrate to the board that cyber risk is being managed rather than merely monitored.

The skill produces the function-by-function gap assessment, the current-versus-target profile in the format auditors and board members expect, an implementation-tier rating, and a prioritized improvement roadmap with effort estimates. Everything needed for an executive briefing or a regulatory inquiry.

"Assess our cybersecurity posture against NIST CSF 2.0. We are a healthcare-adjacent SaaS company. We have strong Protect controls (MFA, patch management, endpoint detection) but weak Govern controls: no formal cybersecurity policy ownership, no risk register, and no board-level security reporting. Build a current profile, a 12-month target profile, and a gap narrative suitable for a board presentation."

Before: CISO spends two weeks building a CSF profile in a spreadsheet, then another week translating it into board language. The resulting presentation is already a month old when it reaches the audit committee.

After: Function-by-function assessment across all six CSF 2.0 functions, a current-versus-target profile, an implementation-tier rating, and a board-ready gap narrative with a 12-month improvement roadmap. Produced in one session.

Setup: 10 minutes. Handles both initial assessments and ongoing annual profile updates. Maps to ISO 27001, CIS Controls, and CMMC for teams managing multiple frameworks simultaneously.

3. PCI DSS v4.0.1 Advisor: CDE Scoping, SAQ Selection, and Gap Analysis

This skill scopes the cardholder data environment, selects the correct Self-Assessment Questionnaire type, runs a requirement-by-requirement gap analysis against PCI DSS v4.0.1, and produces a segmentation and testing plan for QSA assessment.

PCI DSS v4.0.1 introduced requirements that were previously future-dated and are now mandatory. Teams that passed their last QSA under v3.2.1 may have compliance gaps they are unaware of. The SAQ selection mistake is equally common: organizations using a payment redirect assume they qualify for SAQ A when their specific integration may require SAQ A-EP, a dramatically different set of requirements.

The skill begins with CDE scoping (the highest-stakes decision in PCI compliance) before moving to SAQ selection and gap analysis. Getting the scope wrong means either failing an audit on a technicality or implementing controls across systems that do not actually touch card data.

"Scope our PCI DSS cardholder data environment and select the correct SAQ. We use Stripe as our payment processor with a redirect integration. The card data never touches our servers. We do store the last four digits and card brand in our database for display purposes. Our last QSA was under PCI DSS v3.2.1. Run a v4.0.1 gap analysis and identify any newly mandatory requirements we may have missed."

Before: Three rounds of back-and-forth with the QSA to determine CDE scope. Assumed SAQ A applied. QSA determined SAQ A-EP was required based on the JavaScript integration. Remediation cost four months and a reaudit.

After: CDE scope confirmed, correct SAQ identified upfront, requirement-by-requirement v4.0.1 gap report with newly mandatory items flagged, and a segmentation and testing plan ready for QSA review. Before engaging external assessors.

Setup: 10 minutes. Works for merchants and service providers at all levels. Covers the full v4.0.1 requirement set including the previously future-dated requirements now in effect.

4. FedRAMP Authorization Advisor: SSP, POA&M, and ATO Readiness

This skill navigates the full FedRAMP lifecycle: impact level determination, system boundary definition, SSP and SAP/SAR/POA&M documentation, NIST SP 800-53 control mapping, 3PAO assessment preparation, continuous monitoring, and ATO readiness.

FedRAMP is the entry requirement for selling cloud services to the US federal government. The authorization package (System Security Plan, Security Assessment Plan, Security Assessment Report, and Plan of Action and Milestones) can run thousands of pages. Impact level errors and system boundary mistakes are the two most common causes of delayed ATOs, sometimes slipping timelines by quarters.

The skill walks through impact categorization first (Low, Moderate, High), then defines the system boundary with a precision that satisfies 3PAO review, outlines the SSP structure against the applicable 800-53 control baseline, and builds the POA&M framework for tracking open findings. For teams already in the FedRAMP process, it handles continuous monitoring documentation and prepares responses to 3PAO findings.

"Assess our FedRAMP readiness for Moderate impact authorization. We are a cloud-native analytics platform on AWS GovCloud. We have SOC 2 Type II and ISO 27001 in place. Define our system boundary, confirm the impact level, map our existing controls against the Moderate 800-53 baseline, and outline the SSP structure. Flag any controls where our existing documentation will not satisfy FedRAMP requirements."

Before: FedRAMP readiness assessment from a consultancy costs $50,000 to $150,000 and takes months. The deliverable is a gap report but not the SSP, POA&M, or boundary definition. Those are a separate engagement billed separately.

After: Impact level confirmed, system boundary defined, 800-53 Moderate control gap analysis against existing documentation, SSP outline ready for internal review, and POA&M structure set up. Produced internally before engaging a 3PAO.

Setup: 10 minutes. Covers all three impact levels. Handles the full authorization lifecycle from initial readiness through continuous monitoring. Supports teams mid-authorization as well as those just starting.

The Security Framework System: One Source of Truth Across All Frameworks

Security teams pursuing multiple frameworks simultaneously face the same structural problem: every framework has its own gap tracker, its own evidence repository, its own remediation backlog. When a control is implemented, it satisfies requirements in multiple frameworks. But updating four separate trackers is manual, error-prone, and nobody's priority when an incident hits.

The skills above are designed to cross-reference each other. The CIS Controls Advisor produces crosswalks to NIST CSF, ISO 27001, SOC 2, and CMMC. The NIST CSF Advisor maps to CIS Controls and ISO 27001. Running both together on the same control inventory produces a unified view of what you have, what you are missing, and how a single remediation satisfies requirements across multiple frameworks simultaneously.

The workflow for a team pursuing SOC 2, ISO 27001, and PCI DSS simultaneously: a monthly compliance tracker run produces the unified control inventory. The CIS Controls Advisor runs the gap assessment and maps to the other frameworks. The PCI DSS Advisor scopes the CDE and runs the requirement-level gap. Three separate audit preparations collapse into one shared evidence base.

This is the difference between compliance as a recurring sprint and compliance as an operational posture. The framework documentation gets maintained continuously, not assembled from scratch each audit cycle.

Getting Started

I publish all four cybersecurity skills as free, downloadable templates at claudecodehq.com: CIS Controls v8 gap assessment and crosswalks, NIST CSF 2.0 profile building, PCI DSS v4.0.1 scoping and gap analysis, and FedRAMP authorization lifecycle support. Each one is a single file you install once and use whenever you need it. Start with whichever framework is generating the most documentation overhead for your team right now.

Originally published on claudecodehq.com