<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:dc="http://purl.org/dc/elements/1.1/">
  <channel>
    <title>DEV Community: Faris Han</title>
    <description>The latest articles on DEV Community by Faris Han (@farishan).</description>
    <link>https://dev.to/farishan</link>
    <image>
      <url>https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F288400%2F1f429ead-cf94-46e1-a00b-e165f90d8be7.JPG</url>
      <title>DEV Community: Faris Han</title>
      <link>https://dev.to/farishan</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/farishan"/>
    <language>en</language>
    <item>
      <title>Launching smartclean: A CLI to Clean Up Your Dev Machine</title>
      <dc:creator>Faris Han</dc:creator>
      <pubDate>Tue, 17 Nov 2026 16:03:28 +0000</pubDate>
      <link>https://dev.to/farishan/launching-smartclean-a-cli-to-clean-up-your-dev-machine-5bfh</link>
      <guid>https://dev.to/farishan/launching-smartclean-a-cli-to-clean-up-your-dev-machine-5bfh</guid>
      <description>&lt;p&gt;Developer machines fill up quietly. &lt;code&gt;node_modules&lt;/code&gt;, build caches, package manager stores, old tool data, and forgotten Docker layers add up to tens of gigabytes. &lt;code&gt;smartclean&lt;/code&gt; is a small CLI that finds that cruft and removes it safely.&lt;/p&gt;

&lt;h2&gt;
  
  
  What it does
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;Scans common locations for reclaimable space: &lt;code&gt;node_modules&lt;/code&gt;, &lt;code&gt;.cache&lt;/code&gt;, build output, package manager caches.&lt;/li&gt;
&lt;li&gt;Shows what it found with sizes before touching anything.&lt;/li&gt;
&lt;li&gt;Requires confirmation; supports &lt;code&gt;--dry-run&lt;/code&gt; to preview.&lt;/li&gt;
&lt;li&gt;Never deletes a directory that is not safely reproducible; it skips anything it cannot verify.&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  Install
&lt;/h2&gt;



&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;git clone https://github.com/farishan/smartclean.git ~/dev/smartclean
&lt;span class="nb"&gt;cd&lt;/span&gt; ~/dev/smartclean &lt;span class="o"&gt;&amp;amp;&amp;amp;&lt;/span&gt; bash install.sh
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Requires: bash, coreutils, find, du, stat. Optional: &lt;code&gt;mo&lt;/code&gt; (Mole / mole-for-ubuntu) for system-level helpers.&lt;/p&gt;

&lt;h2&gt;
  
  
  Usage
&lt;/h2&gt;



&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;smartclean                      &lt;span class="c"&gt;# full scan, no changes&lt;/span&gt;
smartclean scan --json          &lt;span class="c"&gt;# JSON report&lt;/span&gt;
smartclean scan --deep          &lt;span class="c"&gt;# include deep sweeps&lt;/span&gt;
smartclean scan --keep-days 14  &lt;span class="c"&gt;# protect 14-day-old caches&lt;/span&gt;
smartclean scan --target ~/dev  &lt;span class="c"&gt;# artifacts/large files under ~/dev&lt;/span&gt;
smartclean apply --category artifacts --target ~/dev
smartclean apply --yes          &lt;span class="c"&gt;# apply all (confirms per category unless --yes)&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;The scan output groups by category so you can see where the space went. Most machines recover 5-20 GB on the first run.&lt;/p&gt;

&lt;h2&gt;
  
  
  Safety first
&lt;/h2&gt;

&lt;p&gt;The default is conservative. smartclean will not remove a directory unless it can confirm it is regenerable (lockfile present, cache directory, known build output). If you are unsure, &lt;code&gt;--dry-run&lt;/code&gt; shows exactly what would be deleted.&lt;/p&gt;

&lt;h2&gt;
  
  
  Why I built it
&lt;/h2&gt;

&lt;p&gt;I kept running out of disk on a laptop with a 256 GB SSD. Existing tools either required a GUI or deleted too aggressively. smartclean is the middle ground: fast, terminal-native, and safe.&lt;/p&gt;

&lt;h2&gt;
  
  
  Links
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;Repo: &lt;a href="https://github.com/farishan/smartclean" rel="noopener noreferrer"&gt;farishan/smartclean&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;Releases: &lt;a href="https://github.com/farishan/smartclean/releases" rel="noopener noreferrer"&gt;releases&lt;/a&gt;
&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;Feedback welcome. If it deletes something it should not have, open an issue with the path and I will add a guard.&lt;/p&gt;

&lt;h2&gt;
  
  
  What I learned shipping it
&lt;/h2&gt;

&lt;p&gt;The hardest part was not the scanning; it was deciding what counts as safe. A directory with a lockfile next to it is safe. A directory without one is not, even if it looks like a cache. When in doubt, smartclean skips and tells you why.&lt;/p&gt;

&lt;p&gt;The first version is intentionally small: scan, preview, clean. Future releases will add per-directory allowlists and a JSON output mode for scripting.&lt;/p&gt;

</description>
      <category>cli</category>
      <category>productivity</category>
      <category>opensource</category>
      <category>tooling</category>
    </item>
    <item>
      <title>sweep + doctor: Keep Your OpenCode Project Healthy</title>
      <dc:creator>Faris Han</dc:creator>
      <pubDate>Tue, 17 Nov 2026 13:41:38 +0000</pubDate>
      <link>https://dev.to/farishan/sweep-doctor-keep-your-opencode-project-healthy-3gm0</link>
      <guid>https://dev.to/farishan/sweep-doctor-keep-your-opencode-project-healthy-3gm0</guid>
      <description>&lt;h1&gt;
  
  
  sweep + doctor: Keep Your OpenCode Project Healthy
&lt;/h1&gt;

&lt;p&gt;Agent-built repos rot fast. Dead code accumulates, configs drift, and six weeks&lt;br&gt;
later nobody knows which files are load-bearing. Two small plugins attack this&lt;br&gt;
directly: &lt;code&gt;sweep&lt;/code&gt; finds unused code; &lt;code&gt;doctor&lt;/code&gt; checks config health.&lt;/p&gt;
&lt;h2&gt;
  
  
  sweep
&lt;/h2&gt;

&lt;p&gt;&lt;code&gt;sweep&lt;/code&gt; scans an OpenCode project for unused code: unreferenced functions,&lt;br&gt;
orphaned files, stale imports. It reports what it found with sizes and locations&lt;br&gt;
before touching anything, and supports dry-run. It never deletes what it cannot&lt;br&gt;
verify is unused.&lt;/p&gt;

&lt;p&gt;Use it:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;Before a release, to shrink the surface area.&lt;/li&gt;
&lt;li&gt;After a refactor, to remove leftovers.&lt;/li&gt;
&lt;li&gt;On a schedule, to keep drift visible.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2&gt;
  
  
  doctor
&lt;/h2&gt;

&lt;p&gt;&lt;code&gt;doctor&lt;/code&gt; checks the health of an OpenCode project's configuration: missing or&lt;br&gt;
inconsistent settings, broken references, and common misconfigurations. It&lt;br&gt;
prints a report with pass/warn/fail per check and suggested fixes.&lt;/p&gt;

&lt;p&gt;Use it:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;When onboarding a new repo.&lt;/li&gt;
&lt;li&gt;After changing provider or model config.&lt;/li&gt;
&lt;li&gt;In CI, as a gate.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2&gt;
  
  
  Together
&lt;/h2&gt;

&lt;p&gt;&lt;code&gt;sweep&lt;/code&gt; keeps the codebase small; &lt;code&gt;doctor&lt;/code&gt; keeps the config honest. Run both&lt;br&gt;
before &lt;code&gt;/ship&lt;/code&gt;:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;sweep --dry-run
doctor
wf-verify
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;If all three pass, the repo is in better shape than most.&lt;/p&gt;

&lt;h2&gt;
  
  
  Get it
&lt;/h2&gt;

&lt;p&gt;&lt;code&gt;sweep&lt;/code&gt; and &lt;code&gt;doctor&lt;/code&gt; are OpenCode plugins that ship with the OpenCode Dev&lt;br&gt;
Toolkit (Pro). The core workflow kit is free (MIT).&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;Free bundle: &lt;a href="https://farishan1.gumroad.com/l/kjovp" rel="noopener noreferrer"&gt;https://farishan1.gumroad.com/l/kjovp&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;Pro ($29): &lt;a href="https://farishan1.gumroad.com/l/sknrfm" rel="noopener noreferrer"&gt;https://farishan1.gumroad.com/l/sknrfm&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;GitHub: &lt;a href="https://github.com/farishan/workflow-kit" rel="noopener noreferrer"&gt;https://github.com/farishan/workflow-kit&lt;/a&gt;
&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  Takeaway
&lt;/h2&gt;

&lt;p&gt;Agent-built code does not rot because agents are bad. It rots because nobody&lt;br&gt;
runs the janitor. &lt;code&gt;sweep&lt;/code&gt; and &lt;code&gt;doctor&lt;/code&gt; are the janitor.&lt;/p&gt;

</description>
      <category>opencode</category>
      <category>ai</category>
      <category>devtools</category>
      <category>refactoring</category>
    </item>
    <item>
      <title>wf-verify: One Verify Gate for Every Ecosystem</title>
      <dc:creator>Faris Han</dc:creator>
      <pubDate>Tue, 17 Nov 2026 13:40:46 +0000</pubDate>
      <link>https://dev.to/farishan/wf-verify-one-verify-gate-for-every-ecosystem-2bb0</link>
      <guid>https://dev.to/farishan/wf-verify-one-verify-gate-for-every-ecosystem-2bb0</guid>
      <description>&lt;h1&gt;
  
  
  wf-verify: One Verify Gate for Every Ecosystem
&lt;/h1&gt;

&lt;p&gt;Agentic loops break down at the verify step. The agent writes code, then either&lt;br&gt;
skips verification or runs a command that only covers one ecosystem. A Go repo&lt;br&gt;
gets &lt;code&gt;go test&lt;/code&gt;; a Bun repo gets &lt;code&gt;bun test&lt;/code&gt;; a mixed repo gets nothing consistent.&lt;br&gt;
The result: "it passed" means different things in different projects.&lt;/p&gt;

&lt;p&gt;&lt;code&gt;wf-verify&lt;/code&gt; is one command that detects the project and runs the right gate.&lt;/p&gt;

&lt;h2&gt;
  
  
  What it does
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;Detects project type from files present (&lt;code&gt;go.mod&lt;/code&gt;, &lt;code&gt;package.json&lt;/code&gt;,
&lt;code&gt;Cargo.toml&lt;/code&gt;, &lt;code&gt;pyproject.toml&lt;/code&gt;, &lt;code&gt;Makefile&lt;/code&gt;, and others).&lt;/li&gt;
&lt;li&gt;Runs the appropriate build, lint, and test commands for each detected
ecosystem.&lt;/li&gt;
&lt;li&gt;Returns a single pass/fail exit code the loop can gate on.&lt;/li&gt;
&lt;li&gt;Works in CI and locally with the same invocation.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;Coverage: Go, Bun/Node, Rust, Python, Make, and more (9 ecosystems total).&lt;/p&gt;

&lt;h2&gt;
  
  
  Why a single command matters
&lt;/h2&gt;

&lt;p&gt;In an agentic loop, the verify step is the contract between "implemented" and&lt;br&gt;
"shipped." If the command differs per repo, agents (and humans) improvise, and&lt;br&gt;
improvisation is where regressions slip in. One command means:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;The agent cannot quietly skip a step.&lt;/li&gt;
&lt;li&gt;CI and local agree.&lt;/li&gt;
&lt;li&gt;Reviewers know exactly what "verified" means.&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  Using it in the loop
&lt;/h2&gt;



&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;&lt;span class="c"&gt;# after implementing&lt;/span&gt;
wf-verify
&lt;span class="c"&gt;# exit 0 -&amp;gt; proceed to review&lt;/span&gt;
&lt;span class="c"&gt;# exit non-zero -&amp;gt; fix, do not review or ship&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Make it mandatory in your agent commands: &lt;code&gt;/verify&lt;/code&gt; runs &lt;code&gt;wf-verify&lt;/code&gt; and blocks&lt;br&gt;
&lt;code&gt;/ship&lt;/code&gt; on failure. That single rule removes most "shipped broken code" incidents.&lt;/p&gt;

&lt;h2&gt;
  
  
  Install
&lt;/h2&gt;

&lt;p&gt;&lt;code&gt;wf-verify&lt;/code&gt; ships with the OpenCode Dev Toolkit (Pro). The core workflow kit&lt;br&gt;
(MIT) is free on GitHub.&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;Free bundle: &lt;a href="https://farishan1.gumroad.com/l/kjovp" rel="noopener noreferrer"&gt;https://farishan1.gumroad.com/l/kjovp&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;Pro ($29): &lt;a href="https://farishan1.gumroad.com/l/sknrfm" rel="noopener noreferrer"&gt;https://farishan1.gumroad.com/l/sknrfm&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;GitHub: &lt;a href="https://github.com/farishan/workflow-kit" rel="noopener noreferrer"&gt;https://github.com/farishan/workflow-kit&lt;/a&gt;
&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  Takeaway
&lt;/h2&gt;

&lt;p&gt;Verification is only a gate if it is the same gate every time. One command,&lt;br&gt;
every ecosystem, same result.&lt;/p&gt;

</description>
      <category>opencode</category>
      <category>ai</category>
      <category>devtools</category>
      <category>testing</category>
    </item>
    <item>
      <title>OpenCode Workflow Loop: spec implement verify review ship</title>
      <dc:creator>Faris Han</dc:creator>
      <pubDate>Tue, 17 Nov 2026 13:40:46 +0000</pubDate>
      <link>https://dev.to/farishan/opencode-workflow-loop-spec-implement-verify-review-ship-1dah</link>
      <guid>https://dev.to/farishan/opencode-workflow-loop-spec-implement-verify-review-ship-1dah</guid>
      <description>&lt;h1&gt;
  
  
  OpenCode Workflow Loop: spec → implement → verify → review → ship
&lt;/h1&gt;

&lt;p&gt;Vibe-coding gets you a demo. It rarely gets you a shippable repo. The gap is&lt;br&gt;
not the model; it is the missing structure around the model: a spec to aim at,&lt;br&gt;
a verify gate to catch regressions, a review pass to catch the things tests do&lt;br&gt;
not, and a ship step that does not skip the boring parts.&lt;/p&gt;

&lt;p&gt;This guide walks through the loop I use on every OpenCode project, plus the&lt;br&gt;
templates that make it cheap to repeat.&lt;/p&gt;
&lt;h2&gt;
  
  
  The loop
&lt;/h2&gt;

&lt;ol&gt;
&lt;li&gt;
&lt;strong&gt;Spec&lt;/strong&gt; — write what "done" looks like before touching code.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Implement&lt;/strong&gt; — let the agent write code against the spec.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Verify&lt;/strong&gt; — run a universal verify gate (build, lint, tests).&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Review&lt;/strong&gt; — a second pass: diff review, scope check, security sanity.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Ship&lt;/strong&gt; — commit, tag, push, update docs.&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;Each step is small and gated. You do not move on until the gate passes.&lt;/p&gt;
&lt;h2&gt;
  
  
  1. Spec
&lt;/h2&gt;

&lt;p&gt;A spec is three short files, not a novel:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;code&gt;SPEC.md&lt;/code&gt; — what the thing does, for whom, and the acceptance criteria.&lt;/li&gt;
&lt;li&gt;
&lt;code&gt;PLAN.md&lt;/code&gt; — the ordered steps to get there.&lt;/li&gt;
&lt;li&gt;
&lt;code&gt;ACCEPTANCE.md&lt;/code&gt; — the checklist that proves it shipped.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;Keep each under a page. The point is to give the agent a target it can be&lt;br&gt;
checked against, and to give you a definition of done that survives the session.&lt;/p&gt;

&lt;p&gt;Template (drop into any repo):&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight markdown"&gt;&lt;code&gt;&lt;span class="gu"&gt;# SPEC.md
## Problem
## Users
## Acceptance criteria
&lt;/span&gt;&lt;span class="k"&gt;- &lt;/span&gt;[ ] ...
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;h2&gt;
  
  
  2. Implement
&lt;/h2&gt;

&lt;p&gt;Hand the spec to OpenCode. The agent implements against &lt;code&gt;SPEC.md&lt;/code&gt; and updates&lt;br&gt;
&lt;code&gt;PLAN.md&lt;/code&gt; as it goes. Two rules:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;One task at a time. Do not batch "implement everything."&lt;/li&gt;
&lt;li&gt;The agent must run the verify gate after each meaningful change, not at the end.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2&gt;
  
  
  3. Verify
&lt;/h2&gt;

&lt;p&gt;Verification is the step most people skip, and it is the one that matters.&lt;br&gt;
Use a single command that works across ecosystems:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;wf-verify
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;&lt;code&gt;wf-verify&lt;/code&gt; detects the project type (Go, Bun/Node, Rust, Python, Make, and&lt;br&gt;
more) and runs the right build, lint, and test commands. It returns a clear&lt;br&gt;
pass/fail so the loop can gate on it. No more "it works on my machine" — the&lt;br&gt;
gate is the same everywhere.&lt;/p&gt;

&lt;p&gt;If &lt;code&gt;wf-verify&lt;/code&gt; fails, you do not review or ship. You fix.&lt;/p&gt;

&lt;h2&gt;
  
  
  4. Review
&lt;/h2&gt;

&lt;p&gt;A separate pass, ideally a different agent persona, over the diff:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;Does the diff match &lt;code&gt;SPEC.md&lt;/code&gt;? (scope creep check)&lt;/li&gt;
&lt;li&gt;Are there tests for the new behavior?&lt;/li&gt;
&lt;li&gt;Any secrets, hardcoded paths, or unsafe defaults?&lt;/li&gt;
&lt;li&gt;Did the verify gate actually run, or did someone skip it?&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;Review is cheap and catches the things verify does not.&lt;/p&gt;

&lt;h2&gt;
  
  
  5. Ship
&lt;/h2&gt;

&lt;p&gt;Only after verify + review pass:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;Commit with a clear message (what + why).&lt;/li&gt;
&lt;li&gt;Tag if it is a release.&lt;/li&gt;
&lt;li&gt;Push.&lt;/li&gt;
&lt;li&gt;Update README/install instructions if behavior changed.&lt;/li&gt;
&lt;li&gt;Log the work in your daily inputs.&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  Templates
&lt;/h2&gt;

&lt;p&gt;The loop ships with drop-in templates: &lt;code&gt;SPEC.md&lt;/code&gt;, &lt;code&gt;PLAN.md&lt;/code&gt;, &lt;code&gt;ACCEPTANCE.md&lt;/code&gt;,&lt;br&gt;
plus agent/command definitions (&lt;code&gt;/spec&lt;/code&gt;, &lt;code&gt;/plan&lt;/code&gt;, &lt;code&gt;/verify&lt;/code&gt;, &lt;code&gt;/ship&lt;/code&gt;) and&lt;br&gt;
implementer/tester/reviewer personas. They are MIT-licensed and free.&lt;/p&gt;

&lt;h2&gt;
  
  
  Get it
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;Free bundle (core + prompts): &lt;a href="https://farishan1.gumroad.com/l/kjovp" rel="noopener noreferrer"&gt;https://farishan1.gumroad.com/l/kjovp&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;Pro (full loop + sweep/doctor + verifier): &lt;a href="https://farishan1.gumroad.com/l/sknrfm" rel="noopener noreferrer"&gt;https://farishan1.gumroad.com/l/sknrfm&lt;/a&gt; ($29)&lt;/li&gt;
&lt;li&gt;GitHub (MIT): &lt;a href="https://github.com/farishan/workflow-kit" rel="noopener noreferrer"&gt;https://github.com/farishan/workflow-kit&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;Codes: &lt;code&gt;FIRST10&lt;/code&gt; (50% off first 10), &lt;code&gt;WK20&lt;/code&gt; (20% off), &lt;code&gt;VIP50&lt;/code&gt; (50% off 5)&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  Takeaway
&lt;/h2&gt;

&lt;p&gt;The model is not the bottleneck. The missing gate is. Add a spec, a verify&lt;br&gt;
command, and a review pass, and "vibe-coding" becomes "shipping."&lt;/p&gt;

</description>
      <category>opencode</category>
      <category>ai</category>
      <category>devtools</category>
      <category>productivity</category>
    </item>
    <item>
      <title>Launching smartclean: A CLI to Clean Up Your Dev Machine</title>
      <dc:creator>Faris Han</dc:creator>
      <pubDate>Sat, 14 Nov 2026 02:27:10 +0000</pubDate>
      <link>https://dev.to/farishan/launching-smartclean-a-cli-to-clean-up-your-dev-machine-3k7h</link>
      <guid>https://dev.to/farishan/launching-smartclean-a-cli-to-clean-up-your-dev-machine-3k7h</guid>
      <description>&lt;p&gt;Developer machines fill up quietly. &lt;code&gt;node_modules&lt;/code&gt;, build caches, package manager stores, old tool data, and forgotten Docker layers add up to tens of gigabytes. &lt;code&gt;smartclean&lt;/code&gt; is a small CLI that finds that cruft and removes it safely.&lt;/p&gt;

&lt;h2&gt;
  
  
  What it does
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;Scans common locations for reclaimable space: &lt;code&gt;node_modules&lt;/code&gt;, &lt;code&gt;.cache&lt;/code&gt;, build output, package manager caches.&lt;/li&gt;
&lt;li&gt;Shows what it found with sizes before touching anything.&lt;/li&gt;
&lt;li&gt;Requires confirmation; supports &lt;code&gt;--dry-run&lt;/code&gt; to preview.&lt;/li&gt;
&lt;li&gt;Never deletes a directory that is not safely reproducible; it skips anything it cannot verify.&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  Install
&lt;/h2&gt;



&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;git clone https://github.com/farishan/smartclean.git ~/dev/smartclean
&lt;span class="nb"&gt;cd&lt;/span&gt; ~/dev/smartclean &lt;span class="o"&gt;&amp;amp;&amp;amp;&lt;/span&gt; bash install.sh
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Requires: bash, coreutils, find, du, stat. Optional: &lt;code&gt;mo&lt;/code&gt; (Mole / mole-for-ubuntu) for system-level helpers.&lt;/p&gt;

&lt;h2&gt;
  
  
  Usage
&lt;/h2&gt;



&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;smartclean                      &lt;span class="c"&gt;# full scan, no changes&lt;/span&gt;
smartclean scan --json          &lt;span class="c"&gt;# JSON report&lt;/span&gt;
smartclean scan --deep          &lt;span class="c"&gt;# include deep sweeps&lt;/span&gt;
smartclean scan --keep-days 14  &lt;span class="c"&gt;# protect 14-day-old caches&lt;/span&gt;
smartclean scan --target ~/dev  &lt;span class="c"&gt;# artifacts/large files under ~/dev&lt;/span&gt;
smartclean apply --category artifacts --target ~/dev
smartclean apply --yes          &lt;span class="c"&gt;# apply all (confirms per category unless --yes)&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;The scan output groups by category so you can see where the space went. Most machines recover 5-20 GB on the first run.&lt;/p&gt;

&lt;h2&gt;
  
  
  Safety first
&lt;/h2&gt;

&lt;p&gt;The default is conservative. smartclean will not remove a directory unless it can confirm it is regenerable (lockfile present, cache directory, known build output). If you are unsure, &lt;code&gt;--dry-run&lt;/code&gt; shows exactly what would be deleted.&lt;/p&gt;

&lt;h2&gt;
  
  
  Why I built it
&lt;/h2&gt;

&lt;p&gt;I kept running out of disk on a laptop with a 256 GB SSD. Existing tools either required a GUI or deleted too aggressively. smartclean is the middle ground: fast, terminal-native, and safe.&lt;/p&gt;

&lt;h2&gt;
  
  
  Links
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;Repo: &lt;a href="https://github.com/farishan/smartclean" rel="noopener noreferrer"&gt;farishan/smartclean&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;Releases: &lt;a href="https://github.com/farishan/smartclean/releases" rel="noopener noreferrer"&gt;releases&lt;/a&gt;
&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;Feedback welcome. If it deletes something it should not have, open an issue with the path and I will add a guard.&lt;/p&gt;

&lt;h2&gt;
  
  
  What I learned shipping it
&lt;/h2&gt;

&lt;p&gt;The hardest part was not the scanning; it was deciding what counts as safe. A directory with a lockfile next to it is safe. A directory without one is not, even if it looks like a cache. When in doubt, smartclean skips and tells you why.&lt;/p&gt;

&lt;p&gt;The first version is intentionally small: scan, preview, clean. Future releases will add per-directory allowlists and a JSON output mode for scripting.&lt;/p&gt;

</description>
      <category>cli</category>
      <category>productivity</category>
      <category>opensource</category>
      <category>tooling</category>
    </item>
    <item>
      <title>USER.md for AI Agents: A Portable Operator Profile</title>
      <dc:creator>Faris Han</dc:creator>
      <pubDate>Sat, 14 Nov 2026 02:26:14 +0000</pubDate>
      <link>https://dev.to/farishan/usermd-for-ai-agents-a-portable-operator-profile-1k7b</link>
      <guid>https://dev.to/farishan/usermd-for-ai-agents-a-portable-operator-profile-1k7b</guid>
      <description>&lt;h1&gt;
  
  
  USER.md for AI Agents: A Portable Operator Profile
&lt;/h1&gt;

&lt;p&gt;Every AI tool I use wants to know the same things: who I am, what I am working on, how I like answers formatted, and what to never do. I was repeating myself across Claude, OpenCode, Cursor, and a handful of custom agents. So I extracted it into one file.&lt;/p&gt;

&lt;h2&gt;
  
  
  The idea
&lt;/h2&gt;

&lt;p&gt;A &lt;code&gt;USER.md&lt;/code&gt; (or &lt;code&gt;USER.md&lt;/code&gt;-equivalent) is a portable operator profile. It is not a system prompt and it is not project documentation. It holds &lt;em&gt;tastes&lt;/em&gt;: communication style, priorities, constraints, and preferences. Infra rules live in plugins or AGENTS.md. Machine facts live in a machine profile. Project facts live in the project.&lt;/p&gt;

&lt;p&gt;Keeping those separate is the whole point. When tastes are mixed into infra rules, every new machine or project re-derives them.&lt;/p&gt;

&lt;h2&gt;
  
  
  What goes in
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;Identity and role (operator, not persona)&lt;/li&gt;
&lt;li&gt;Current priorities and active projects&lt;/li&gt;
&lt;li&gt;Communication preferences (terse vs detailed, language)&lt;/li&gt;
&lt;li&gt;Hard constraints ("never force push to main", "ask before publishing")&lt;/li&gt;
&lt;li&gt;Things I have already decided (so the agent does not re-litigate them)&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  What stays out
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;Tool configuration and hooks (plugins/AGENTS.md)&lt;/li&gt;
&lt;li&gt;Machine-specific paths (machine profile)&lt;/li&gt;
&lt;li&gt;Project architecture (project AGENTS.md)&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  A template
&lt;/h2&gt;

&lt;p&gt;I published a template at &lt;a href="https://github.com/farishan-labs/user-md" rel="noopener noreferrer"&gt;farishan-labs/user-md&lt;/a&gt; that you can use as a starting point. It is intentionally small: around one to one-and-a-half thousand tokens. Bigger profiles get ignored or truncated.&lt;/p&gt;

&lt;h2&gt;
  
  
  Demand, not hype
&lt;/h2&gt;

&lt;p&gt;I held off on a paid product until I saw real demand. The signal I watch: how often I (or others) re-explain the same preferences to a new agent. Three or more signals means the problem is real. So far the signals are there, so Phase 2 (a Gumroad listing) is in prep.&lt;/p&gt;

&lt;h2&gt;
  
  
  Try it
&lt;/h2&gt;

&lt;ol&gt;
&lt;li&gt;Create &lt;code&gt;USER.md&lt;/code&gt; in your home or project root.&lt;/li&gt;
&lt;li&gt;Add five lines: who you are, what you are building, how you want answers, one constraint, one preference.&lt;/li&gt;
&lt;li&gt;Point your agent at it in the system prompt or AGENTS.md.&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;The first time an agent stops asking "should I use tabs or spaces?" you will know it worked.&lt;/p&gt;

</description>
      <category>ai</category>
      <category>productivity</category>
      <category>agents</category>
      <category>opensource</category>
    </item>
    <item>
      <title>Repo Quality Gate Walkthrough: A Two-Layer Gate for Shipping Small Projects</title>
      <dc:creator>Faris Han</dc:creator>
      <pubDate>Sat, 14 Nov 2026 02:26:13 +0000</pubDate>
      <link>https://dev.to/farishan/repo-quality-gate-walkthrough-a-two-layer-gate-for-shipping-small-projects-in9</link>
      <guid>https://dev.to/farishan/repo-quality-gate-walkthrough-a-two-layer-gate-for-shipping-small-projects-in9</guid>
      <description>&lt;h1&gt;
  
  
  Repo Quality Gate Walkthrough: A Two-Layer Gate for Shipping Small Projects
&lt;/h1&gt;

&lt;p&gt;Small repos die from neglect, not from bad code. The failure mode is familiar: you ship v0.1, move on, and six months later the README is wrong, CI is red, and the license file is missing. I wanted a gate I could run in thirty seconds before calling something "done".&lt;/p&gt;

&lt;p&gt;This post walks through the two-layer gate I now use across my projects, plus a credibility layer borrowed from research.&lt;/p&gt;

&lt;h2&gt;
  
  
  Layer 1: Health percentage
&lt;/h2&gt;

&lt;p&gt;The first layer is a simple health percentage: does the repo have the boring stuff?&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;README with install + usage&lt;/li&gt;
&lt;li&gt;LICENSE&lt;/li&gt;
&lt;li&gt;CI workflow that runs real tests&lt;/li&gt;
&lt;li&gt;CONTRIBUTING or at least a note&lt;/li&gt;
&lt;li&gt;.gitignore&lt;/li&gt;
&lt;li&gt;Release or version tag&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;I count the present items over the total and require a floor (I use 90%). A repo at 70% is not "almost done"; it is missing the things that make it trustworthy. The health check is cheap and catches the embarrassing gaps.&lt;/p&gt;

&lt;h2&gt;
  
  
  Layer 2: CLI quality score
&lt;/h2&gt;

&lt;p&gt;For command-line tools, the second layer is a score from a reusable checker (&lt;code&gt;launchlint&lt;/code&gt;-style): argument parsing, help text, exit codes, error messages, and tests. I require a score of at least 70% before publishing.&lt;/p&gt;

&lt;p&gt;The two layers are independent on purpose. A repo can have a great README and a broken CLI, or a polished CLI and no license. Both must pass.&lt;/p&gt;

&lt;h2&gt;
  
  
  Layer 3 (credibility): research backing
&lt;/h2&gt;

&lt;p&gt;Numbers alone do not convince anyone. The credibility layer asks: is there a public document explaining &lt;em&gt;why&lt;/em&gt; these thresholds? For my projects I keep a research topic with sources and a measurement methodology, then inline a short summary in the README.&lt;/p&gt;

&lt;p&gt;That is the difference between "I picked 70%" and "research on repo quality measurement supports a 70% floor plus a health gate".&lt;/p&gt;

&lt;h2&gt;
  
  
  The reusable script
&lt;/h2&gt;

&lt;p&gt;I extracted the gate into a reusable script in my agent repo. Run it against any repo:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;python3 tools/repo-quality-gate.py --path ~/dev/labs/your-repo
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;It prints the health percentage, the CLI score if applicable, and a pass/fail verdict. CI calls the same script, so local and CI agree.&lt;/p&gt;

&lt;h2&gt;
  
  
  What this changed
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;Fewer "v0.1" repos that are actually v0.0.&lt;/li&gt;
&lt;li&gt;A consistent bar across projects: agent, user-md, smartclean all pass the same gate.&lt;/li&gt;
&lt;li&gt;Review time dropped, because the gate answers "is this shippable?" before I read the code.&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  Try it
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;Agent repo: &lt;a href="https://github.com/farishan-labs/agent" rel="noopener noreferrer"&gt;farishan-labs/agent&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;Research backing: &lt;a href="https://github.com/farishan-labs/research" rel="noopener noreferrer"&gt;repo-quality-measurement&lt;/a&gt;
&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;If you have a small project sitting at 70%, run the gate and fix the two missing items. It is usually the license and the CI.&lt;/p&gt;

</description>
      <category>productivity</category>
      <category>opensource</category>
      <category>devtools</category>
      <category>github</category>
    </item>
    <item>
      <title>Teaching your AI agent who you are: a portable USER.md template</title>
      <dc:creator>Faris Han</dc:creator>
      <pubDate>Wed, 11 Nov 2026 16:03:35 +0000</pubDate>
      <link>https://dev.to/farishan/teaching-your-ai-agent-who-you-are-a-portable-usermd-template-11il</link>
      <guid>https://dev.to/farishan/teaching-your-ai-agent-who-you-are-a-portable-usermd-template-11il</guid>
      <description>&lt;p&gt;AI agents are stateless by default. Every session starts from zero unless you give them a stable profile of the human operator: role, preferences, communication style, and constraints.&lt;/p&gt;

&lt;p&gt;&lt;code&gt;USER.md&lt;/code&gt; is the convention (originating in the OpenClaw workspace) for a small, durable file that holds &lt;strong&gt;who the human operator is&lt;/strong&gt; — separate from the agent's persona (&lt;code&gt;SOUL.md&lt;/code&gt;) and from project instructions (&lt;code&gt;AGENTS.md&lt;/code&gt;).&lt;/p&gt;

&lt;p&gt;I published a portable, harness-agnostic template for it: &lt;strong&gt;&lt;a href="https://github.com/farishan-labs/user-md" rel="noopener noreferrer"&gt;farishan-labs/user-md&lt;/a&gt;&lt;/strong&gt; (MIT).&lt;/p&gt;

&lt;h2&gt;
  
  
  What's in the repo
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;A clean, ready-to-fill &lt;code&gt;USER.md&lt;/code&gt; template.&lt;/li&gt;
&lt;li&gt;
&lt;code&gt;GUIDE.md&lt;/code&gt; with the format spec, best practices, anti-patterns, and migration paths.&lt;/li&gt;
&lt;li&gt;
&lt;code&gt;WORKSHEET.md&lt;/code&gt; to seed your profile in 5 minutes.&lt;/li&gt;
&lt;li&gt;Filled examples for OpenClaw, Hermes, Claude Code (global), and OptMem.&lt;/li&gt;
&lt;li&gt;A standalone validator CLI, &lt;code&gt;usermd-check&lt;/code&gt;.&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  Quick start
&lt;/h2&gt;



&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;go install github.com/farishan-labs/user-md/checker@latest
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Or grab a pre-built binary from the &lt;a href="https://github.com/farishan-labs/user-md/releases" rel="noopener noreferrer"&gt;releases page&lt;/a&gt;.&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;Copy &lt;code&gt;USER.md&lt;/code&gt; to your agent's workspace root (or &lt;code&gt;~/.hermes/memories/USER.md&lt;/code&gt;, or &lt;code&gt;~/.claude/CLAUDE.md&lt;/code&gt; global).&lt;/li&gt;
&lt;li&gt;Fill in &lt;code&gt;WORKSHEET.md&lt;/code&gt; (5 questions) and paste the resulting directives into &lt;code&gt;USER.md&lt;/code&gt;.&lt;/li&gt;
&lt;li&gt;Keep it under your harness's cap (Hermes: ~1,375 chars; OpenClaw: 4,000 chars).&lt;/li&gt;
&lt;li&gt;Audit monthly.&lt;/li&gt;
&lt;/ol&gt;

&lt;h2&gt;
  
  
  The three-file split
&lt;/h2&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;File&lt;/th&gt;
&lt;th&gt;What it holds&lt;/th&gt;
&lt;th&gt;Who writes it&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;&lt;code&gt;SOUL.md&lt;/code&gt;&lt;/td&gt;
&lt;td&gt;Agent's persona, voice, values&lt;/td&gt;
&lt;td&gt;You (or the agent, agent-scoped)&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;code&gt;USER.md&lt;/code&gt;&lt;/td&gt;
&lt;td&gt;
&lt;strong&gt;Your&lt;/strong&gt; profile, preferences, constraints&lt;/td&gt;
&lt;td&gt;You (seeded), agent (curated)&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;code&gt;AGENTS.md&lt;/code&gt;&lt;/td&gt;
&lt;td&gt;Project instructions, conventions&lt;/td&gt;
&lt;td&gt;You / team&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;code&gt;MEMORY.md&lt;/code&gt;&lt;/td&gt;
&lt;td&gt;Durable non-profile facts and decisions&lt;/td&gt;
&lt;td&gt;Agent&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;Rule: &lt;code&gt;SOUL.md&lt;/code&gt; and &lt;code&gt;USER.md&lt;/code&gt; are separate systems that never feed each other.&lt;/p&gt;

&lt;h2&gt;
  
  
  Format (OpenClaw canonical)
&lt;/h2&gt;



&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight markdown"&gt;&lt;code&gt;&amp;lt;!-- observed: YYYY-MM-DD | status: active --&amp;gt;

&lt;span class="k"&gt;- &lt;/span&gt;Prefer concise progress updates during implementation work.
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;ul&gt;
&lt;li&gt;One directive per entry, imperative voice (Always / Never / Prefer).&lt;/li&gt;
&lt;li&gt;Record the date the preference was &lt;strong&gt;observed&lt;/strong&gt;, not invented.&lt;/li&gt;
&lt;li&gt;Status is only &lt;code&gt;active&lt;/code&gt; or &lt;code&gt;superseded&lt;/code&gt;.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Supersede in place&lt;/strong&gt; — never append contradictions.&lt;/li&gt;
&lt;li&gt;Keep it compact; it's not a dossier.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;See &lt;a href="https://github.com/farishan-labs/user-md/blob/main/GUIDE.md" rel="noopener noreferrer"&gt;&lt;code&gt;GUIDE.md&lt;/code&gt;&lt;/a&gt; for the full spec.&lt;/p&gt;

&lt;h2&gt;
  
  
  Compatibility
&lt;/h2&gt;

&lt;p&gt;Works with any harness that loads a user-profile file or global instructions:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;OpenClaw&lt;/strong&gt; — &lt;code&gt;USER.md&lt;/code&gt; at workspace root (4,000-char budget)&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Hermes Agent&lt;/strong&gt; — &lt;code&gt;~/.hermes/memories/USER.md&lt;/code&gt; (~1,375 chars)&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Claude Code&lt;/strong&gt; — global &lt;code&gt;~/.claude/CLAUDE.md&lt;/code&gt; (layered)&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Codex / Copilot / Aider / Cursor&lt;/strong&gt; — via &lt;code&gt;AGENTS.md&lt;/code&gt; or global rules&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;OptMem&lt;/strong&gt; — as a portable operator-profile note&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  Why this exists / research backing
&lt;/h2&gt;

&lt;p&gt;This template comes from a survey of how AI agents recognize their human operator across OpenClaw, Hermes, Claude Code, Cursor, Aider, Codex, Continue, Cline, Roo, Windsurf, Goose, and memory frameworks (Letta, Mem0, Zep, Honcho). Full research notes (evidence ledger, 18 sources, ecosystem survey) are maintained privately; the public sources that ground each claim are cited below.&lt;/p&gt;

&lt;p&gt;Key findings:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Three separate problems:&lt;/strong&gt; identity (who is operating), personalization (preferences), authorization (what the agent may do). Most coding agents solve only ambient identity (OS user / git author) plus project instructions. &lt;em&gt;Sources: &lt;a href="https://labs.cloudsecurityalliance.org/research/csa-research-note-nist-ai-agent-standards-initiative-2026040/" rel="noopener noreferrer"&gt;CSA NIST AI agent standards note&lt;/a&gt;, &lt;a href="https://arxiv.org/html/2608.00966v1" rel="noopener noreferrer"&gt;AgenTag attribution paper&lt;/a&gt;.&lt;/em&gt;
&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Dominant transferable pattern:&lt;/strong&gt; the &lt;strong&gt;SOUL.md vs USER.md split&lt;/strong&gt; — the agent's own persona (&lt;code&gt;SOUL.md&lt;/code&gt;) is kept strictly separate from the human's profile (&lt;code&gt;USER.md&lt;/code&gt;) and from project instructions (&lt;code&gt;AGENTS.md&lt;/code&gt;). OpenClaw docs state they "never feed each other." &lt;em&gt;Sources: &lt;a href="https://clawdocs.org/guides/soul-md" rel="noopener noreferrer"&gt;OpenClaw SOUL.md guide&lt;/a&gt;, &lt;a href="https://docs.openclaw.ai/concepts/memory" rel="noopener noreferrer"&gt;OpenClaw memory overview&lt;/a&gt;, &lt;a href="https://docs.openclaw.ai/reference/templates/USER" rel="noopener noreferrer"&gt;OpenClaw USER template&lt;/a&gt;, &lt;a href="https://hermes-agent.nousresearch.com/docs/guides/migrate-from-openclaw" rel="noopener noreferrer"&gt;Hermes migration from OpenClaw&lt;/a&gt;.&lt;/em&gt;
&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Coding-agent convergence:&lt;/strong&gt; &lt;code&gt;AGENTS.md&lt;/code&gt; is the cross-tool constitution, with tool-specific layers (&lt;code&gt;CLAUDE.md&lt;/code&gt; global+project, &lt;code&gt;.cursor/rules/*.mdc&lt;/code&gt;). Personal preference lives in global instructions, not in project files. &lt;em&gt;Sources: &lt;a href="https://automatelab.tech/blog/ai-coding/claude-md-agents-md-cursorrules" rel="noopener noreferrer"&gt;AGENTS.md vs CLAUDE.md vs .cursorrules (automatelab)&lt;/a&gt;, &lt;a href="https://stackitfast.com/insights/agents-md-vs-claude-md-vs-cursorrules" rel="noopener noreferrer"&gt;AGENTS.md vs CLAUDE.md vs .cursorrules (stackitfast)&lt;/a&gt;, &lt;a href="https://docs.openclaw.ai/reference/templates/AGENTS" rel="noopener noreferrer"&gt;OpenClaw AGENTS.md template&lt;/a&gt;.&lt;/em&gt;
&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Memory ≠ identity:&lt;/strong&gt; memory systems (Letta, Mem0, Zep, Honcho) recognize users via a stable &lt;code&gt;user_id&lt;/code&gt; (usually from an auth provider) and retrieve scoped memories; they don't replace an identity file. &lt;em&gt;Sources: &lt;a href="https://docs.letta.com/letta-agent/memory" rel="noopener noreferrer"&gt;Letta memory docs&lt;/a&gt;, &lt;a href="https://mem0.ai/guide/memory-layer-user-aware-ai-agents" rel="noopener noreferrer"&gt;Mem0 memory layer guide&lt;/a&gt;, &lt;a href="https://www.remoteopenclaw.com/blog/hermes-agent-persistent-memory-methods" rel="noopener noreferrer"&gt;Hermes memory (remoteopenclaw)&lt;/a&gt;, &lt;a href="https://mem0.ai/blog/build-an-ai-agent-that-remembers-your-users" rel="noopener noreferrer"&gt;Mem0 build an agent that remembers users&lt;/a&gt;.&lt;/em&gt;
&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Recommendation:&lt;/strong&gt; keep the role/identity separation, and add one small, portable operator profile file (this &lt;code&gt;USER.md&lt;/code&gt;) consolidating rules scattered across &lt;code&gt;AGENTS.md&lt;/code&gt; files, with OptMem as cross-harness memory. &lt;em&gt;Sources: &lt;a href="https://hermes-agent.nousresearch.com/docs/guides/tips" rel="noopener noreferrer"&gt;Hermes tips &amp;amp; best practices&lt;/a&gt;, &lt;a href="https://www.stanza.dev/courses/hermes-memory/user-profile/hermes-memory-profile-what" rel="noopener noreferrer"&gt;Hermes user profiling (Stanza)&lt;/a&gt;.&lt;/em&gt;
&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Security:&lt;/strong&gt; keep commit author (agent slot) distinct from push auth (human); never store secrets in identity files. &lt;em&gt;Sources: &lt;a href="https://arxiv.org/html/2608.00966v1" rel="noopener noreferrer"&gt;AgenTag attribution paper&lt;/a&gt;, &lt;a href="https://labs.cloudsecurityalliance.org/research/csa-research-note-nist-ai-agent-standards-initiative-2026040/" rel="noopener noreferrer"&gt;CSA NIST AI agent standards note&lt;/a&gt;.&lt;/em&gt;
&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;The core claim: agents are stateless by default, and a small, curated, &lt;em&gt;separate&lt;/em&gt; profile file is more reliable than stuffing preferences into &lt;code&gt;AGENTS.md&lt;/code&gt; or hoping the agent remembers.&lt;/p&gt;

</description>
      <category>ai</category>
      <category>agents</category>
      <category>productivity</category>
      <category>opensource</category>
    </item>
    <item>
      <title>JavaScript RPG Document: Exploration Feature - Level 1</title>
      <dc:creator>Faris Han</dc:creator>
      <pubDate>Fri, 22 Nov 2024 15:04:49 +0000</pubDate>
      <link>https://dev.to/farishan/javascript-rpg-document-exploration-feature-level-1-46h</link>
      <guid>https://dev.to/farishan/javascript-rpg-document-exploration-feature-level-1-46h</guid>
      <description>&lt;p&gt;The easiest exploration feature I can think of is 1-dimensional exploration. Up-down, or left-right, or point A to point B, or room 1 to room 2.&lt;/p&gt;

&lt;p&gt;Let's choose the "room" one, like in MUD games.&lt;/p&gt;

&lt;h2&gt;
  
  
  Requirements
&lt;/h2&gt;

&lt;h3&gt;
  
  
  Functional Requirements
&lt;/h3&gt;

&lt;ul&gt;
&lt;li&gt;Player can see their character's current room&lt;/li&gt;
&lt;li&gt;Player can move to next rooms by pressing "d" key&lt;/li&gt;
&lt;li&gt;Player can move to previous rooms by pressing "a" key&lt;/li&gt;
&lt;li&gt;There is only 3 rooms and player can only explore that 3 rooms&lt;/li&gt;
&lt;li&gt;Player can see if they can not go to the next/previous room&lt;/li&gt;
&lt;/ul&gt;

&lt;h3&gt;
  
  
  Technical Requirements
&lt;/h3&gt;

&lt;ul&gt;
&lt;li&gt;The toy can be played just by double-clicking an HTML file&lt;/li&gt;
&lt;/ul&gt;

&lt;h3&gt;
  
  
  Implementation
&lt;/h3&gt;

&lt;p&gt;&lt;iframe src="https://codesandbox.io/embed/3rhll8?view=preview&amp;amp;module=%2Fsrc%2Findex.html" allow="geolocation; microphone; camera; midi; vr; accelerometer; gyroscope; payment; ambient-light-sensor; encrypted-media; usb" sandbox="allow-modals allow-forms allow-popups allow-scripts allow-same-origin"&gt;
&lt;/iframe&gt;
&lt;/p&gt;

</description>
      <category>javascript</category>
      <category>rpg</category>
      <category>gamedev</category>
      <category>beginners</category>
    </item>
    <item>
      <title>Apa itu Autentikasi: Definisi dan Jenis-jenis Autentikasi</title>
      <dc:creator>Faris Han</dc:creator>
      <pubDate>Sun, 06 Aug 2023 06:22:26 +0000</pubDate>
      <link>https://dev.to/farishan/apa-itu-autentikasi-definisi-dan-jenis-jenis-autentikasi-288e</link>
      <guid>https://dev.to/farishan/apa-itu-autentikasi-definisi-dan-jenis-jenis-autentikasi-288e</guid>
      <description>&lt;p&gt;Dalam dunia digital, sering terdengar istilah "autentikasi" atau "authentication". Tulisan ini akan membahas sedikit lebih dalam tentang itu. Selamat membaca!&lt;/p&gt;

&lt;h2&gt;
  
  
  Apa itu Autentikasi?
&lt;/h2&gt;

&lt;p&gt;"Autentikasi" (baku) atau "otentikasi" (tidak baku) itu proses untuk membuktikan sesuatu secara autentik (baku) atau otentik (tidak baku).&lt;/p&gt;

&lt;p&gt;Autentik sendiri artinya "bisa dipercaya".&lt;/p&gt;

&lt;p&gt;Dalam bahasa Inggris disebut &lt;em&gt;authentication&lt;/em&gt; atau disingkat menjadi &lt;em&gt;auth&lt;/em&gt; saja. Untuk selanjutnya, saya akan terus menggunakan istilah "auth".&lt;/p&gt;

&lt;p&gt;Jadi kalau ada pertanyaan "apa itu auth?", artinya sama dengan autentikasi yang sudah disebut di atas.&lt;/p&gt;

&lt;p&gt;Sumber:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;&lt;a href="https://kbbi.kemdikbud.go.id/entri/autentikasi" rel="noopener noreferrer"&gt;https://kbbi.kemdikbud.go.id/entri/autentikasi&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://kbbi.kemdikbud.go.id/entri/autentik" rel="noopener noreferrer"&gt;https://kbbi.kemdikbud.go.id/entri/autentik&lt;/a&gt;&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  Kapan Proses Auth Digunakan?
&lt;/h2&gt;

&lt;p&gt;Ketika program atau aplikasi komputer yang kita buat butuh tahu penggunanya (Inggris: user) bisa dipercaya atau tidak. Sering sekali di dalam aplikasi ada fitur-fitur yang tidak bisa digunakan oleh sembarang orang, maka proses auth ini sangat populer. Bahkan banyak modul terpisah yang memang dibuat khusus untuk proses autentikasi.&lt;/p&gt;

&lt;p&gt;"Pengguna" itu sendiri belum tentu manusia, ya. Bisa jadi komputer atau entitas lain.&lt;/p&gt;

&lt;h2&gt;
  
  
  Auth Sehari-hari
&lt;/h2&gt;

&lt;p&gt;Biasanya orang awam lebih familiar dengan istilah "login" dan "logout". Dua hal tersebut termasuk dalam lingkup auth. Selain itu ada juga istilah "register" atau "daftar" atau "buat akun". Proses pembuatan akun ini juga masuk dalam lingkup auth.&lt;/p&gt;

&lt;h2&gt;
  
  
  Jenis-jenis Auth
&lt;/h2&gt;

&lt;p&gt;Proses auth ada berbagai macam tingkat kompleksitas dan keamanan. Secara umum, bisa dibagi menjadi beberapa kategori di bawah ini:&lt;/p&gt;

&lt;h3&gt;
  
  
  1. Knowledge-based Authentication
&lt;/h3&gt;

&lt;p&gt;Atau "autentikasi berbasis pengetahuan". Auth ini berhubungan dengan sesuatu yang diketahui user, seperti: password, pin, atau jawaban atas pertanyaan keamanan.&lt;/p&gt;

&lt;h3&gt;
  
  
  2. Posession-based Authentication
&lt;/h3&gt;

&lt;p&gt;Atau "autentikasi berbasis kepemilikan". Auth ini berhubungan dengan sesuatu yang dimiliki user, seperti: kartu, token fisik, atau mobile device.&lt;/p&gt;

&lt;h3&gt;
  
  
  3. Biometric Authentication
&lt;/h3&gt;

&lt;p&gt;Auth ini berhubungan dengan ciri-ciri fisik yang dimiliki user manusia. Proses auth bisa dilakukan dengan mendeteksi wajah, sidik jari, atau bagian tubuh lain.&lt;/p&gt;

&lt;h3&gt;
  
  
  4. Multi-factor Authentication (MFA)
&lt;/h3&gt;

&lt;p&gt;Auth ini kombinasi atau gabungan dari jenis-jenis auth yang sudah disebutkan sebelumnya. &lt;/p&gt;

&lt;h2&gt;
  
  
  Kesimpulan
&lt;/h2&gt;

&lt;p&gt;Auth atau autentikasi itu proses untuk membuktikan sesuatu bisa dipercaya atau tidak.&lt;/p&gt;

&lt;p&gt;Auth adalah pintu dari banyak proses lainnya yang berhubungan dengan kontrol akses. Semoga dengan tulisan ini, kita bisa lebih memahami proses lain yang lebih kompleks.&lt;/p&gt;

&lt;p&gt;Sila meninggalkan komentar atau pertanyaan untuk memulai diskusi tentang auth.&lt;/p&gt;

&lt;p&gt;Terima kasih sudah membaca!&lt;/p&gt;

</description>
      <category>indonesia</category>
      <category>auth</category>
    </item>
    <item>
      <title>Apa itu Programmer?</title>
      <dc:creator>Faris Han</dc:creator>
      <pubDate>Sat, 29 Jul 2023 07:39:37 +0000</pubDate>
      <link>https://dev.to/farishan/apa-itu-programmer-8k1</link>
      <guid>https://dev.to/farishan/apa-itu-programmer-8k1</guid>
      <description>&lt;p&gt;TL;DR&lt;br&gt;
Programmer itu orang yang membuat program. Bisa program komputer atau program acara televisi.&lt;/p&gt;




&lt;p&gt;Pada saat tulisan ini dibuat (29 Juli, 2023), di Kamus Besar Bahasa Indonesia (KBBI) tidak ada arti untuk kata "programmer".&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Farticles%2Fq73dn58jqd9r57g85vm1.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Farticles%2Fq73dn58jqd9r57g85vm1.png" alt="pencarian di KBBI untuk kata " width="800" height="239"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;Karena "programmer" itu kosa kata dari bahasa Inggris. Di kamus Merriam-Webster (kamus yang paling dipercaya di Amerika), ada arti untuk kata "&lt;a href="https://www.merriam-webster.com/dictionary/programmer" rel="noopener noreferrer"&gt;programmer&lt;/a&gt;".&lt;/p&gt;

&lt;p&gt;Untuk bahasa Indonesia, ada kata "pemrogram" dengan arti yang sama.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Farticles%2Feibpjofo46rgg3xj2m3x.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Farticles%2Feibpjofo46rgg3xj2m3x.png" alt=" " width="800" height="364"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  Kesimpulan
&lt;/h2&gt;

&lt;p&gt;Programmer atau pemrogram adalah orang yang membuat program, baik itu program komputer atau program acara televisi. Dalam konteks situs dev.to, "programmer" lebih sering mengacu kepada pembuat program komputer. Program komputer adalah program yang diciptakan secara khusus sehingga memungkinkan komputer melakukan fungsi tertentu.&lt;/p&gt;

</description>
      <category>indonesia</category>
    </item>
    <item>
      <title>Apa itu HTML</title>
      <dc:creator>Faris Han</dc:creator>
      <pubDate>Sun, 02 Jul 2023 15:02:32 +0000</pubDate>
      <link>https://dev.to/farishan/apa-itu-html-ecb</link>
      <guid>https://dev.to/farishan/apa-itu-html-ecb</guid>
      <description>&lt;p&gt;Fakta: HTML itu singkatan dari Hypertext Markup Language.&lt;/p&gt;

&lt;p&gt;HTML itu bahasa markup yang digunakan untuk membuat dan menyusun halaman web.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://dev.to/farishan/apa-itu-bahasa-markup-3e67"&gt;Klik di sini untuk membaca tentang apa itu bahasa markup.&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;HTML memungkinkan pengguna untuk menggambarkan struktur dan konten halaman web dengan menggunakan elemen-elemen markup. Elemen-elemen ini ditandai dengan tag-tag khusus yang ditempatkan di sekitar teks, gambar, atau media lainnya dalam halaman web.&lt;/p&gt;

&lt;h2&gt;
  
  
  Struktur, seperti Tulang Manusia dan Kerangka Bangunan
&lt;/h2&gt;

&lt;p&gt;HTML bisa dibayangkan seperti tulang yang menjadi struktur tubuh manusia dan kerangka yang menjadi struktur bangunan.&lt;br&gt;
HTML itu struktur halaman website.&lt;/p&gt;

&lt;h2&gt;
  
  
  Elemen-elemen di HTML
&lt;/h2&gt;

&lt;p&gt;Setiap halaman web terdiri dari elemen-elemen HTML yang membentuk struktur dasar halaman. Beberapa elemen/tag dasar HTML termasuk:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;tag &lt;code&gt;&amp;lt;html&amp;gt;&lt;/code&gt; sebagai kontainer utama&lt;/li&gt;
&lt;li&gt;tag &lt;code&gt;&amp;lt;head&amp;gt;&lt;/code&gt; untuk informasi kepala dokumen seperti judul halaman&lt;/li&gt;
&lt;li&gt;tag &lt;code&gt;&amp;lt;body&amp;gt;&lt;/code&gt; untuk konten utama halaman, dan&lt;/li&gt;
&lt;li&gt;tag &lt;code&gt;&amp;lt;p&amp;gt;&lt;/code&gt; untuk paragraf teks.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;Kata "tag" itu biasanya digunakan untuk menunjukkan kodenya, misal tag &lt;code&gt;&amp;lt;html&amp;gt;&lt;/code&gt;, tag &lt;code&gt;&amp;lt;body&amp;gt;&lt;/code&gt;, atau tag &lt;code&gt;&amp;lt;p&amp;gt;&lt;/code&gt;. Sedangkan kata "elemen" itu biasanya digunakan untuk menunjukkan hasil dari penggunaan tagnya. Misal elemen body, elemen paragraf.&lt;/p&gt;

&lt;p&gt;Dalam banyak kasus, penggunaan kata "tag" atau "elemen" tidak terlalu jadi masalah. Pemahaman mendalam tentang perbedaan kata "tag" dan "elemen" akan datang dengan sendirinya seiring berjalannya penggunaan HTML di kegiatan sehari-hari.&lt;/p&gt;

&lt;h2&gt;
  
  
  Kesimpulan
&lt;/h2&gt;

&lt;p&gt;HTML itu bahasa markup yang digunakan untuk membuat struktur halaman web.&lt;/p&gt;




&lt;p&gt;Masih punya pertanyaan terkait apa itu HTML? Tulis di kolom komentar ya. Terima kasih sudah membaca!&lt;/p&gt;

</description>
    </item>
  </channel>
</rss>
