<?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: Theodore Creates Co.</title>
    <description>The latest articles on DEV Community by Theodore Creates Co. (@theodorecreatesco).</description>
    <link>https://dev.to/theodorecreatesco</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%2F4219535%2F9cc4c569-cbdc-42f0-8617-77fe3d61edfa.png</url>
      <title>DEV Community: Theodore Creates Co.</title>
      <link>https://dev.to/theodorecreatesco</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/theodorecreatesco"/>
    <language>en</language>
    <item>
      <title>How to track scholarship deadlines without losing one</title>
      <dc:creator>Theodore Creates Co.</dc:creator>
      <pubDate>Wed, 11 Nov 2026 11:09:05 +0000</pubDate>
      <link>https://dev.to/theodorecreatesco/how-to-track-scholarship-deadlines-without-losing-one-5f6i</link>
      <guid>https://dev.to/theodorecreatesco/how-to-track-scholarship-deadlines-without-losing-one-5f6i</guid>
      <description>&lt;p&gt;Scholarship season fails in quiet ways: a deadline you meant to remember, a recommender you asked too late, an essay you rewrote from scratch when you already had 80% of it. You do not need a complex system. You need one place that answers "what is due next?" every time you open it.&lt;/p&gt;

&lt;h2&gt;
  
  
  The four lists that actually matter
&lt;/h2&gt;

&lt;p&gt;Most people start a giant spreadsheet and abandon it. Keep four short lists instead (they can live as tabs in one file):&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;
&lt;strong&gt;Scholarships&lt;/strong&gt; — name, amount, deadline, status, link, notes.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Checklist per scholarship&lt;/strong&gt; — essay, transcript, FAFSA/CSS, recommendations, portal submit.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Recommenders&lt;/strong&gt; — who you asked, what they need, when you reminded them, whether the letter landed.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Essay bank&lt;/strong&gt; — each essay you have written, tagged by theme (leadership, community, STEM why, financial need…).&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;If a scholarship is not on list 1 with a real deadline, it does not exist. If a recommender is not on list 3, assume the letter is not coming.&lt;/p&gt;

&lt;h2&gt;
  
  
  Countdowns beat calendars
&lt;/h2&gt;

&lt;p&gt;A calendar full of coloured blocks looks organised and still gets ignored. A &lt;strong&gt;countdown&lt;/strong&gt; next to each open scholarship ("12 days", "3 days", "due today") forces a decision: work it, ask for an extension, or drop it.&lt;/p&gt;

&lt;p&gt;Sort the active list by deadline ascending. Your "due next" row should be obvious without filters. Update status the same day you submit — "Submitted" with the date — so you never re-open a finished portal by mistake.&lt;/p&gt;

&lt;h2&gt;
  
  
  Reuse essays on purpose
&lt;/h2&gt;

&lt;p&gt;Writing a new essay for every prompt is how people burn out in March. Keep an essay bank and a &lt;strong&gt;prompt reuse map&lt;/strong&gt;:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;Prompt → which bank essay is closest&lt;/li&gt;
&lt;li&gt;What 10–20% you must rewrite for this school&lt;/li&gt;
&lt;li&gt;Word count limit&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;Most "unique" prompts are variations of the same five stories. Draft the stories once; tailor the last mile.&lt;/p&gt;

&lt;h2&gt;
  
  
  Do not skip the aid calendar
&lt;/h2&gt;

&lt;p&gt;Scholarships sit on top of federal and state aid dates. Missing FAFSA (or your state's equivalent deadline) can cost more than any single scholarship. Keep a small &lt;strong&gt;aid timeline&lt;/strong&gt; with:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;FAFSA open / your state deadline&lt;/li&gt;
&lt;li&gt;CSS Profile if your colleges need it&lt;/li&gt;
&lt;li&gt;Verification / document requests&lt;/li&gt;
&lt;li&gt;Award-letter comparison window&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;State FAFSA deadlines differ. If you are applying across states, put the earliest relevant date at the top of the list — not the federal "priority" rumour from a blog post three years ago.&lt;/p&gt;

&lt;h2&gt;
  
  
  A weekly 20-minute review
&lt;/h2&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Step&lt;/th&gt;
&lt;th&gt;Minutes&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;Sort by deadline; pick this week's top 3&lt;/td&gt;
&lt;td&gt;3&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Chase any recommender past due&lt;/td&gt;
&lt;td&gt;5&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Advance one essay (new or reuse)&lt;/td&gt;
&lt;td&gt;10&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Log submissions and update statuses&lt;/td&gt;
&lt;td&gt;2&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;That is enough to keep momentum without living in the spreadsheet.&lt;/p&gt;

&lt;h2&gt;
  
  
  A ready-made version if you want one
&lt;/h2&gt;

&lt;p&gt;I built a &lt;strong&gt;&lt;a href="https://www.etsy.com/uk/listing/4607596744" rel="noopener noreferrer"&gt;Scholarship Tracker 2027&lt;/a&gt;&lt;/strong&gt; with those tabs plus state FAFSA deadlines, a cost-vs-aid college comparison and a fillable essay planner. Use it, or copy the structure into a blank sheet — the workflow above is the useful part either way.&lt;/p&gt;

&lt;p&gt;There is not a separate free sample repo for the tracker yet (unlike the DnD session sheet). If you only take one thing from this post, take the four lists and the weekly review.&lt;/p&gt;

&lt;h2&gt;
  
  
  Starter checklist
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;[ ] Every scholarship has a deadline and a status&lt;/li&gt;
&lt;li&gt;[ ] Recommenders have names, ask dates and reminder dates&lt;/li&gt;
&lt;li&gt;[ ] Essay bank has at least three reusable stories&lt;/li&gt;
&lt;li&gt;[ ] FAFSA / state aid date is on the timeline&lt;/li&gt;
&lt;li&gt;[ ] You know the single next deadline&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;Open the list, do the next thing, close it. That beats a perfect system you never touch.&lt;/p&gt;

</description>
      <category>scholarships</category>
      <category>college</category>
      <category>productivity</category>
      <category>career</category>
    </item>
    <item>
      <title>Plan your next DnD session in 10 minutes (free fillable sheet)</title>
      <dc:creator>Theodore Creates Co.</dc:creator>
      <pubDate>Mon, 09 Nov 2026 10:48:36 +0000</pubDate>
      <link>https://dev.to/theodorecreatesco/plan-your-next-dnd-session-in-10-minutes-free-fillable-sheet-4e9c</link>
      <guid>https://dev.to/theodorecreatesco/plan-your-next-dnd-session-in-10-minutes-free-fillable-sheet-4e9c</guid>
      <description>&lt;p&gt;Most session-prep advice assumes you have a free afternoon. Real life usually gives you ten minutes between work and the players arriving. Here is a short workflow that still produces a session you can run, plus a free two-page sheet that holds it.&lt;/p&gt;

&lt;h2&gt;
  
  
  What "good enough" prep looks like
&lt;/h2&gt;

&lt;p&gt;You do not need a novel. You need five things:&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;
&lt;strong&gt;What happened last time&lt;/strong&gt; — one or two lines the players will recognise.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;A session goal&lt;/strong&gt; — the question tonight answers ("Will they trust the lighthouse keeper?").&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;A strong start&lt;/strong&gt; — the first 60 seconds already in motion.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Three scenes&lt;/strong&gt; — not a railroad, just three places play is likely to go.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Eight secrets and clues&lt;/strong&gt; — rumour, object, NPC line, map mark. Scatter them; the players will find some.&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;Everything else (full NPC sheets, boxed text, contingency trees) is optional. If you only finish those five, you can improvise the rest.&lt;/p&gt;

&lt;h2&gt;
  
  
  The two-page sheet
&lt;/h2&gt;

&lt;p&gt;I keep a free &lt;strong&gt;&lt;a href="https://github.com/theodorecreatesco/free-dnd-session-prep-sheet" rel="noopener noreferrer"&gt;D&amp;amp;D Session Prep Sheet&lt;/a&gt;&lt;/strong&gt; on GitHub: fillable and print PDFs in A4 and US Letter.&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Page 1 — Session prep:&lt;/strong&gt; players, last time, session goal, strong start, three scenes, secrets &amp;amp; clues, NPCs, locations, monsters, loot, after-session notes.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Page 2 — Initiative tracker:&lt;/strong&gt; round boxes, a 12-row tracker (initiative, AC, HP, damage, conditions), a quick DC guide, cover rules and next-session hooks.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;Open the fillable PDF in Acrobat or a tablet PDF app so the fields work. Print page 2 if you prefer paper at the table.&lt;/p&gt;

&lt;h2&gt;
  
  
  A 10-minute routine
&lt;/h2&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Minute&lt;/th&gt;
&lt;th&gt;Do this&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;0–2&lt;/td&gt;
&lt;td&gt;Skim last session's notes. Write "last time" and tonight's goal.&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;2–4&lt;/td&gt;
&lt;td&gt;Write the strong start as a present-tense sentence.&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;4–7&lt;/td&gt;
&lt;td&gt;Name three scenes. Add one complication to each.&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;7–9&lt;/td&gt;
&lt;td&gt;List 5–8 secrets/clues and which scene they live in.&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;9–10&lt;/td&gt;
&lt;td&gt;Jot HP/AC for likely fights on the initiative page. Leave blank rows for surprises.&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;After the session, spend two minutes on the "after" box: what changed, what the players care about now, and one hook for next time. Future-you will thank present-you.&lt;/p&gt;

&lt;h2&gt;
  
  
  When you need more than two pages
&lt;/h2&gt;

&lt;p&gt;The free sheet is the session layer. For a longer campaign you also want NPC continuity, quests, loot and a place to park world notes between sessions. That is what the paid &lt;strong&gt;&lt;a href="https://www.etsy.com/uk/listing/4602333949" rel="noopener noreferrer"&gt;DnD Campaign Planner&lt;/a&gt;&lt;/strong&gt; is for (20 fillable pages, A4 and US Letter, with a filled example adventure). Same brand, optional upgrade — the GitHub sheet stays free either way.&lt;/p&gt;

&lt;h2&gt;
  
  
  Quick checklist before you open Discord / sit down
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;[ ] Goal fits in one sentence&lt;/li&gt;
&lt;li&gt;[ ] Strong start does not wait for the players to "go somewhere"&lt;/li&gt;
&lt;li&gt;[ ] At least three clues exist outside the "correct" path&lt;/li&gt;
&lt;li&gt;[ ] Initiative page has blank rows&lt;/li&gt;
&lt;li&gt;[ ] You know what "done for tonight" looks like&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;That is the whole method. Prep the spine, leave room to play, and write down what mattered before you forget.&lt;/p&gt;

</description>
      <category>dnd</category>
      <category>productivity</category>
      <category>ttrpg</category>
      <category>opensource</category>
    </item>
    <item>
      <title>How I stop AI coding agents breaking my project</title>
      <dc:creator>Theodore Creates Co.</dc:creator>
      <pubDate>Sun, 08 Nov 2026 11:10:36 +0000</pubDate>
      <link>https://dev.to/theodorecreatesco/how-i-stop-ai-coding-agents-breaking-my-project-9i7</link>
      <guid>https://dev.to/theodorecreatesco/how-i-stop-ai-coding-agents-breaking-my-project-9i7</guid>
      <description>&lt;p&gt;I like AI coding agents. I do not like when they rewrite a file that was already fine, hard-code a key "just for now", or try the same broken fix five times. The fix was not a smarter model. It was a short set of house rules the agent reads at the start of every session.&lt;/p&gt;

&lt;h2&gt;
  
  
  What kept going wrong
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Wholesale rewrites&lt;/strong&gt; of files that only needed a small change.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Scope creep&lt;/strong&gt;: renaming and reformatting unrelated code.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Secrets&lt;/strong&gt; in client bundles or committed &lt;code&gt;.env&lt;/code&gt; files.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Missing auth&lt;/strong&gt; on new routes that "worked" in the happy path.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Debug spirals&lt;/strong&gt;: five near-identical attempts, each worse than the last.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;None of that is mysterious. Without priorities, "make it work" quietly beats "don't break security" and "stay in scope".&lt;/p&gt;

&lt;h2&gt;
  
  
  The rules that helped
&lt;/h2&gt;

&lt;h3&gt;
  
  
  1. Rule priority
&lt;/h3&gt;

&lt;p&gt;When instructions clash, follow the higher one and say which you followed:&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;Security guardrails — never broken, even if asked.&lt;/li&gt;
&lt;li&gt;The task I asked for, in scope.&lt;/li&gt;
&lt;li&gt;Stack conventions.&lt;/li&gt;
&lt;li&gt;Working style preferences.&lt;/li&gt;
&lt;/ol&gt;

&lt;h3&gt;
  
  
  2. How to work
&lt;/h3&gt;

&lt;ul&gt;
&lt;li&gt;Read the files you will change (and their callers) first.&lt;/li&gt;
&lt;li&gt;Make the smallest change; never rewrite a working file wholesale.&lt;/li&gt;
&lt;li&gt;Ambiguous request: ask one short question, then proceed.&lt;/li&gt;
&lt;li&gt;Two failed attempts at the same bug: stop, summarise what you tried, ask before a third.&lt;/li&gt;
&lt;/ul&gt;

&lt;h3&gt;
  
  
  3. Security always on
&lt;/h3&gt;

&lt;ul&gt;
&lt;li&gt;Secrets only in server-side env.&lt;/li&gt;
&lt;li&gt;Check the caller on sensitive paths.&lt;/li&gt;
&lt;li&gt;No destructive git, &lt;code&gt;rm -rf&lt;/code&gt;, or &lt;code&gt;--force&lt;/code&gt; unless explicitly requested.&lt;/li&gt;
&lt;li&gt;Run the real tests; do not skip them to go green.&lt;/li&gt;
&lt;/ul&gt;

&lt;h3&gt;
  
  
  4. What to report when done
&lt;/h3&gt;

&lt;p&gt;What changed, what you ran, and what you did &lt;strong&gt;not&lt;/strong&gt; verify. That last line is the one that saves evenings.&lt;/p&gt;

&lt;h2&gt;
  
  
  Where to put the rules
&lt;/h2&gt;

&lt;p&gt;Put the shared text in &lt;strong&gt;&lt;code&gt;AGENTS.md&lt;/code&gt;&lt;/strong&gt;. Point &lt;strong&gt;&lt;code&gt;CLAUDE.md&lt;/code&gt;&lt;/strong&gt; at it with &lt;code&gt;@AGENTS.md&lt;/code&gt;. Add a thin always-on Cursor rule under &lt;code&gt;.cursor/rules/&lt;/code&gt; if you want the security lines loaded in every Cursor chat.&lt;/p&gt;

&lt;p&gt;A free MIT starter with those files, a 10-point pre-launch checklist and three prompts is here:&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;&lt;a href="https://github.com/theodorecreatesco/vibe-coding-rules-starter" rel="noopener noreferrer"&gt;github.com/theodorecreatesco/vibe-coding-rules-starter&lt;/a&gt;&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;Copy the three files into a project, open a new chat, and ask what the top rules are. If the agent cannot name the priority and the security lines, the files are in the wrong place for that tool.&lt;/p&gt;

&lt;p&gt;The optional paid upgrade with more stacks and a one-command installer is the &lt;a href="https://www.etsy.com/uk/listing/4601953941" rel="noopener noreferrer"&gt;Vibe Coding Rules Kit on Etsy&lt;/a&gt;. I use the free starter on small repos and the kit when a project spans several tools.&lt;/p&gt;

&lt;h2&gt;
  
  
  What I do not do any more
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;Paste a 400-line "constitution" into every tool's file.&lt;/li&gt;
&lt;li&gt;Trust "done" without seeing the commands that were run.&lt;/li&gt;
&lt;li&gt;Let the agent take a third blind swing at the same bug.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;Short rules, ranked once, read every session. That is the whole method.&lt;/p&gt;

</description>
      <category>ai</category>
      <category>cursor</category>
      <category>claude</category>
      <category>productivity</category>
    </item>
    <item>
      <title>A pre-commit checklist for AI-written code</title>
      <dc:creator>Theodore Creates Co.</dc:creator>
      <pubDate>Sat, 07 Nov 2026 11:10:46 +0000</pubDate>
      <link>https://dev.to/theodorecreatesco/a-pre-commit-checklist-for-ai-written-code-4li1</link>
      <guid>https://dev.to/theodorecreatesco/a-pre-commit-checklist-for-ai-written-code-4li1</guid>
      <description>&lt;p&gt;AI coding agents are fast. They are also confident when they should not be. Before you commit a change an agent wrote, walk this list. It takes a few minutes. Skipping it is how API keys, missing auth checks and "tidied" files land on &lt;code&gt;main&lt;/code&gt;.&lt;/p&gt;

&lt;h2&gt;
  
  
  1. Secrets and public env
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;No API keys, tokens or passwords in source, comments or test fixtures.&lt;/li&gt;
&lt;li&gt;No secrets in &lt;code&gt;NEXT_PUBLIC_&lt;/code&gt;, &lt;code&gt;VITE_&lt;/code&gt;, &lt;code&gt;REACT_APP_&lt;/code&gt; or other client-exposed prefixes.&lt;/li&gt;
&lt;li&gt;
&lt;code&gt;.env&lt;/code&gt; files that hold secrets are gitignored; only &lt;code&gt;.env.example&lt;/code&gt; with placeholders is tracked.&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  2. Auth and tenancy
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;Every new or changed API route / server action checks the caller.&lt;/li&gt;
&lt;li&gt;Record access is scoped to the signed-in user (or their org). A working 200 for the wrong user is a bug, not a feature.&lt;/li&gt;
&lt;li&gt;Failed auth returns 401/403/404 as your project already does — do not invent a new pattern mid-PR.&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  3. Scope creep
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;Diff only touches files needed for the request.&lt;/li&gt;
&lt;li&gt;No drive-by renames, reformats or "while I was here" refactors of working code.&lt;/li&gt;
&lt;li&gt;No new dependencies unless you asked for them (and you know the license).&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  4. Behaviour and tests
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;You can explain the change in one or two sentences.&lt;/li&gt;
&lt;li&gt;Existing tests still run; new behaviour has a test if your repo expects one.&lt;/li&gt;
&lt;li&gt;The agent did not "fix" a failing test by deleting or skipping it.&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  5. Commands you actually ran
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;Install / migrate / seed only if required — and you know what they did.&lt;/li&gt;
&lt;li&gt;Lint and the project's usual test command passed on your machine.&lt;/li&gt;
&lt;li&gt;App still boots locally for the path you changed.&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  6. Git hygiene
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;Commit message describes the user-facing change, not "update files".&lt;/li&gt;
&lt;li&gt;No &lt;code&gt;--force&lt;/code&gt; to shared branches, no rewriting published history, no committing &lt;code&gt;node_modules&lt;/code&gt; or build artefacts.&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  A short version you can paste into AGENTS.md
&lt;/h2&gt;



&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight markdown"&gt;&lt;code&gt;&lt;span class="gu"&gt;## Before you say done&lt;/span&gt;
&lt;span class="p"&gt;-&lt;/span&gt; No secrets in source or client env prefixes.
&lt;span class="p"&gt;-&lt;/span&gt; Auth checked on every sensitive path; data scoped to the caller.
&lt;span class="p"&gt;-&lt;/span&gt; Smallest diff; no unrelated refactors or new deps unless asked.
&lt;span class="p"&gt;-&lt;/span&gt; Real tests run; do not skip or delete failing tests to go green.
&lt;span class="p"&gt;-&lt;/span&gt; Report what you changed, what you ran, and what you did not verify.
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;That block is in the free &lt;a href="https://github.com/theodorecreatesco/vibe-coding-rules-starter" rel="noopener noreferrer"&gt;vibe-coding-rules-starter&lt;/a&gt; (&lt;code&gt;AGENTS.md&lt;/code&gt; plus a 10-point security checklist under &lt;code&gt;checklists/&lt;/code&gt;). Copy it into your repo so the agent sees the bar before it claims finished.&lt;/p&gt;

&lt;p&gt;If you want the fuller kit (stack packs, installer, more prompts), it lives on Etsy as the &lt;a href="https://www.etsy.com/uk/listing/4601953941" rel="noopener noreferrer"&gt;Vibe Coding Rules Kit&lt;/a&gt;. The checklist above is enough to stop most of the expensive mistakes on its own.&lt;/p&gt;

&lt;h2&gt;
  
  
  Habit that helps
&lt;/h2&gt;

&lt;p&gt;Treat "the agent said it works" as a claim, not a result. Your pre-commit checklist is the review. Two minutes now beats an incident later.&lt;/p&gt;

</description>
      <category>ai</category>
      <category>git</category>
      <category>security</category>
      <category>productivity</category>
    </item>
    <item>
      <title>AGENTS.md vs .cursorrules vs CLAUDE.md: what goes where</title>
      <dc:creator>Theodore Creates Co.</dc:creator>
      <pubDate>Fri, 06 Nov 2026 10:20:36 +0000</pubDate>
      <link>https://dev.to/theodorecreatesco/agentsmd-vs-cursorrules-vs-claudemd-what-goes-where-22jh</link>
      <guid>https://dev.to/theodorecreatesco/agentsmd-vs-cursorrules-vs-claudemd-what-goes-where-22jh</guid>
      <description>&lt;p&gt;If you use more than one AI coding tool, you have probably ended up with three nearly identical instruction files and no idea which one actually loaded. Here is the short version of what each file is for, and a setup that stops the drift.&lt;/p&gt;

&lt;h2&gt;
  
  
  The three files in one table
&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;Who reads it&lt;/th&gt;
&lt;th&gt;Best use&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;&lt;code&gt;AGENTS.md&lt;/code&gt;&lt;/td&gt;
&lt;td&gt;Cursor, GitHub Copilot coding agent, Codex, Windsurf, Zed, Amp and a growing list of others&lt;/td&gt;
&lt;td&gt;Shared project brief: how to work, security, commands, what "done" means&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;code&gt;.cursor/rules/*.mdc&lt;/code&gt;&lt;/td&gt;
&lt;td&gt;Cursor only&lt;/td&gt;
&lt;td&gt;Narrow rules that should load for certain globs (or always)&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;code&gt;CLAUDE.md&lt;/code&gt;&lt;/td&gt;
&lt;td&gt;Claude Code&lt;/td&gt;
&lt;td&gt;Claude-specific extras; can import &lt;code&gt;AGENTS.md&lt;/code&gt; with &lt;code&gt;@AGENTS.md&lt;/code&gt;
&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;&lt;code&gt;.cursorrules&lt;/code&gt; (single file at the repo root) is the older Cursor format. It still works in many setups, but Cursor's current project-rules model is &lt;code&gt;.cursor/rules/*.mdc&lt;/code&gt; with front matter for &lt;code&gt;description&lt;/code&gt;, &lt;code&gt;globs&lt;/code&gt; and &lt;code&gt;alwaysApply&lt;/code&gt;.&lt;/p&gt;

&lt;p&gt;&lt;code&gt;AGENTS.md&lt;/code&gt; is intentionally boring Markdown. There is no special schema. Agents just read the text. That is why it works as the cross-tool home for rules you do not want to rewrite three times.&lt;/p&gt;

&lt;h2&gt;
  
  
  A setup that actually holds
&lt;/h2&gt;

&lt;ol&gt;
&lt;li&gt;Put the real rules in &lt;strong&gt;&lt;code&gt;AGENTS.md&lt;/code&gt;&lt;/strong&gt; at the repo root: rule priority, security guardrails, how to make changes, which commands to run, what to report when finished.&lt;/li&gt;
&lt;li&gt;Make &lt;strong&gt;&lt;code&gt;CLAUDE.md&lt;/code&gt;&lt;/strong&gt; thin. Start it with &lt;code&gt;@AGENTS.md&lt;/code&gt;, then add only Claude-specific notes if you need them.&lt;/li&gt;
&lt;li&gt;Add &lt;strong&gt;&lt;code&gt;.cursor/rules/&lt;/code&gt;&lt;/strong&gt; only for things that should attach to certain files (for example a Next.js App Router rule with &lt;code&gt;globs: ["app/**/*.tsx"]&lt;/code&gt;), plus one short always-on security rule if you want the guardrails in every Cursor chat even when &lt;code&gt;AGENTS.md&lt;/code&gt; is long.&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;Do not paste the same 200-line manifesto into all three. One source of truth beats three copies that disagree by Thursday.&lt;/p&gt;

&lt;h2&gt;
  
  
  What belongs in the shared file
&lt;/h2&gt;

&lt;p&gt;Keep &lt;code&gt;AGENTS.md&lt;/code&gt; short enough that a model will still follow it:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Rule priority&lt;/strong&gt; when instructions clash (security beats "just make it work").&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Hard security lines&lt;/strong&gt;: secrets stay in server env, auth checks on every sensitive path, no destructive git or force-push unless asked.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Working style&lt;/strong&gt;: read before editing, smallest change, stay in scope, run the real tests.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Project commands&lt;/strong&gt;: install, test, lint, start — the ones a new teammate would need.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;Leave stack fashion tips and long architecture essays for docs humans read, or for glob-scoped &lt;code&gt;.mdc&lt;/code&gt; files.&lt;/p&gt;

&lt;h2&gt;
  
  
  A free starter you can copy
&lt;/h2&gt;

&lt;p&gt;I keep a free MIT starter with a ready &lt;code&gt;AGENTS.md&lt;/code&gt;, a &lt;code&gt;CLAUDE.md&lt;/code&gt; that imports it, an always-on Cursor security rule, a 10-point pre-launch checklist and three copy-paste prompts:&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;&lt;a href="https://github.com/theodorecreatesco/vibe-coding-rules-starter" rel="noopener noreferrer"&gt;theodorecreatesco/vibe-coding-rules-starter&lt;/a&gt;&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;Clone it, copy the three files into your project, and ask a fresh chat: &lt;em&gt;"What are the top rules in this project?"&lt;/em&gt; If it cannot name the priority and the security lines, the files are not where the tool expects them.&lt;/p&gt;

&lt;p&gt;If you later want stack packs, more prompts and a one-command installer for several tools, that is the paid &lt;a href="https://www.etsy.com/uk/listing/4601953941" rel="noopener noreferrer"&gt;Vibe Coding Rules Kit on Etsy&lt;/a&gt; — same brand, optional upgrade. The starter on GitHub stays free either way.&lt;/p&gt;

&lt;h2&gt;
  
  
  Quick decision guide
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;One tool, Cursor only, path-specific rules:&lt;/strong&gt; &lt;code&gt;.cursor/rules/*.mdc&lt;/code&gt; is enough.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;One tool, Claude Code only:&lt;/strong&gt; &lt;code&gt;CLAUDE.md&lt;/code&gt; is enough.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Mixed tools, open source repo, or you are tired of syncing:&lt;/strong&gt; &lt;code&gt;AGENTS.md&lt;/code&gt; as the source of truth, thin pointers elsewhere.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;That is the whole map. Pick the shared file once, keep it short, and stop rewriting the same rules under three names.&lt;/p&gt;

</description>
      <category>ai</category>
      <category>cursor</category>
      <category>productivity</category>
      <category>opensource</category>
    </item>
    <item>
      <title>How to write Cursor rules your AI will actually follow</title>
      <dc:creator>Theodore Creates Co.</dc:creator>
      <pubDate>Fri, 06 Nov 2026 07:50:19 +0000</pubDate>
      <link>https://dev.to/theodorecreatesco/how-to-write-cursor-rules-your-ai-will-actually-follow-lem</link>
      <guid>https://dev.to/theodorecreatesco/how-to-write-cursor-rules-your-ai-will-actually-follow-lem</guid>
      <description>&lt;p&gt;A rules file that the model never loads is just a README. A rules file that &lt;em&gt;always&lt;/em&gt; loads, and tells it three conflicting things, is worse. Most "my AI ignored the rules" stories are one of those two problems.&lt;/p&gt;

&lt;p&gt;This is a short craft guide: how to write Cursor rules (and the same ideas for Claude Code) so they stick. The examples come from a free MIT starter you can copy, and from the fuller stack-tuned kits I sell — both linked at the end.&lt;/p&gt;

&lt;h2&gt;
  
  
  1. Decide what is always on
&lt;/h2&gt;

&lt;p&gt;Cursor project rules are Markdown files in &lt;code&gt;.cursor/rules/&lt;/code&gt; with a tiny YAML header. Two switches matter:&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="nn"&gt;---&lt;/span&gt;
&lt;span class="na"&gt;description&lt;/span&gt;&lt;span class="pi"&gt;:&lt;/span&gt; &lt;span class="s2"&gt;"&lt;/span&gt;&lt;span class="s"&gt;Rule&lt;/span&gt;&lt;span class="nv"&gt; &lt;/span&gt;&lt;span class="s"&gt;priority&lt;/span&gt;&lt;span class="nv"&gt; &lt;/span&gt;&lt;span class="s"&gt;and&lt;/span&gt;&lt;span class="nv"&gt; &lt;/span&gt;&lt;span class="s"&gt;security&lt;/span&gt;&lt;span class="nv"&gt; &lt;/span&gt;&lt;span class="s"&gt;guardrails&lt;/span&gt;&lt;span class="nv"&gt; &lt;/span&gt;&lt;span class="s"&gt;that&lt;/span&gt;&lt;span class="nv"&gt; &lt;/span&gt;&lt;span class="s"&gt;apply&lt;/span&gt;&lt;span class="nv"&gt; &lt;/span&gt;&lt;span class="s"&gt;to&lt;/span&gt;&lt;span class="nv"&gt; &lt;/span&gt;&lt;span class="s"&gt;every&lt;/span&gt;&lt;span class="nv"&gt; &lt;/span&gt;&lt;span class="s"&gt;task"&lt;/span&gt;
&lt;span class="na"&gt;alwaysApply&lt;/span&gt;&lt;span class="pi"&gt;:&lt;/span&gt; &lt;span class="kc"&gt;true&lt;/span&gt;
&lt;span class="nn"&gt;---&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;&lt;code&gt;alwaysApply: true&lt;/code&gt; means the rule is in every chat, even if the agent never opens the matching files. Use it for things that must never be negotiated:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;rule priority (what wins when instructions clash)&lt;/li&gt;
&lt;li&gt;security (secrets, auth checks, no force-push, no skipping tests)&lt;/li&gt;
&lt;li&gt;"ask before installing packages / deleting files"&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;Everything else should be &lt;strong&gt;off&lt;/strong&gt; by default and tied to files with &lt;code&gt;globs&lt;/code&gt;:&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="nn"&gt;---&lt;/span&gt;
&lt;span class="na"&gt;description&lt;/span&gt;&lt;span class="pi"&gt;:&lt;/span&gt; &lt;span class="s"&gt;Rules for adding, removing or upgrading dependencies&lt;/span&gt;
&lt;span class="na"&gt;globs&lt;/span&gt;&lt;span class="pi"&gt;:&lt;/span&gt; &lt;span class="err"&gt;**&lt;/span&gt;&lt;span class="s"&gt;/package.json&lt;/span&gt;
&lt;span class="na"&gt;alwaysApply&lt;/span&gt;&lt;span class="pi"&gt;:&lt;/span&gt; &lt;span class="kc"&gt;false&lt;/span&gt;
&lt;span class="nn"&gt;---&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;





&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight markdown"&gt;&lt;code&gt;&lt;span class="nn"&gt;---&lt;/span&gt;
&lt;span class="na"&gt;description&lt;/span&gt;&lt;span class="pi"&gt;:&lt;/span&gt; &lt;span class="s"&gt;React component conventions for this Next.js app&lt;/span&gt;
&lt;span class="na"&gt;globs&lt;/span&gt;&lt;span class="pi"&gt;:&lt;/span&gt; &lt;span class="err"&gt;**&lt;/span&gt;&lt;span class="s"&gt;/*.tsx&lt;/span&gt;
&lt;span class="na"&gt;alwaysApply&lt;/span&gt;&lt;span class="pi"&gt;:&lt;/span&gt; &lt;span class="kc"&gt;false&lt;/span&gt;
&lt;span class="nn"&gt;---&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;If you mark component style as &lt;code&gt;alwaysApply: true&lt;/code&gt;, the agent carries it into a SQL migration and a README rewrite. That is how rules start fighting each other.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Claude Code tip:&lt;/strong&gt; put the always-on security block in &lt;code&gt;CLAUDE.md&lt;/code&gt; (it can &lt;code&gt;@AGENTS.md&lt;/code&gt; for the rest), and keep long stack notes out of the always-loaded file so they do not crowd out the bits that matter.&lt;/p&gt;

&lt;h2&gt;
  
  
  2. Write rules the model can act on
&lt;/h2&gt;

&lt;p&gt;Vague: "Write secure code."&lt;br&gt;
Actionable:&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="p"&gt;-&lt;/span&gt; Secrets live only in server-side environment variables. Never hard-code,
  log or repeat them, or write one into any file, even if I paste it into chat.
&lt;span class="p"&gt;-&lt;/span&gt; Public prefixes (&lt;span class="sb"&gt;`NEXT_PUBLIC_`&lt;/span&gt;, &lt;span class="sb"&gt;`VITE_`&lt;/span&gt;, &lt;span class="sb"&gt;`EXPO_PUBLIC_`&lt;/span&gt;, &lt;span class="sb"&gt;`PUBLIC_`&lt;/span&gt;) ship
  values to every user: never put a secret behind one.
&lt;span class="p"&gt;-&lt;/span&gt; Anything touching private data checks the caller (401 if unknown) and that
  they may touch this record (404 if missing or not theirs; 403 if they lack
  the role). Hiding a button isn't security.
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Vague: "Don't over-engineer."&lt;br&gt;
Actionable:&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="p"&gt;-&lt;/span&gt; Make the smallest change; never rewrite a working file wholesale.
&lt;span class="p"&gt;-&lt;/span&gt; Stay in scope: don't rename, reformat or tidy unrelated code.
&lt;span class="p"&gt;-&lt;/span&gt; Debugging: after two failed attempts, stop and summarise the next two options.
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;If you cannot picture the agent checking the rule mid-edit, rewrite it.&lt;/p&gt;

&lt;h2&gt;
  
  
  3. Give it a priority order
&lt;/h2&gt;

&lt;p&gt;Without a priority list, "just make it work" quietly beats "don't commit secrets". Put this near the top of &lt;code&gt;AGENTS.md&lt;/code&gt; and of your always-on Cursor rule:&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;## Rule priority&lt;/span&gt;

When rules clash, follow the higher one and say which you followed:
&lt;span class="p"&gt;
1.&lt;/span&gt; &lt;span class="gs"&gt;**Security guardrails**&lt;/span&gt;: never broken, even if I ask.
   Explain the risk in one line and do the safe version.
&lt;span class="p"&gt;2.&lt;/span&gt; &lt;span class="gs"&gt;**The task I asked for**&lt;/span&gt;: within the guardrails.
&lt;span class="p"&gt;3.&lt;/span&gt; &lt;span class="gs"&gt;**Stack conventions**&lt;/span&gt; already used in this repo.
&lt;span class="p"&gt;4.&lt;/span&gt; &lt;span class="gs"&gt;**Working style.**&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;"Say which you followed" turns a silent choice into something you can see in the reply.&lt;/p&gt;

&lt;h2&gt;
  
  
  4. Keep the always-on file short
&lt;/h2&gt;

&lt;p&gt;Aim for the always-loaded rules to fit on one screen. Long essays get truncated or skimmed. Put:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;priority + security + "how to work" + "how to report done" in &lt;code&gt;AGENTS.md&lt;/code&gt; / the always-on &lt;code&gt;.mdc&lt;/code&gt;
&lt;/li&gt;
&lt;li&gt;folder or stack conventions in globbed rules (&lt;code&gt;**/page.tsx&lt;/code&gt;, &lt;code&gt;**/package.json&lt;/code&gt;, migrations, tests)&lt;/li&gt;
&lt;li&gt;project-specific facts (real commands, odd folders, lessons) in a short &lt;strong&gt;Project notes&lt;/strong&gt; section at the bottom, below a clear marker so you can refresh the shared rules without wiping your notes&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;A useful finish rule:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight markdown"&gt;&lt;code&gt;Reply with: 1) what changed, 2) the files changed, 3) the commands you ran
and their results, 4) how I can test it, 5) "Needs your OK": anything I
should confirm, plus anything risky or unfinished.
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Point 3 is the one that stops "everything should work now".&lt;/p&gt;

&lt;h2&gt;
  
  
  5. The 10-second test
&lt;/h2&gt;

&lt;p&gt;After you add or edit rules, start a &lt;strong&gt;new&lt;/strong&gt; chat and ask:&lt;/p&gt;

&lt;blockquote&gt;
&lt;p&gt;What are the top rules in this project? Quote the priority order and one security rule.&lt;/p&gt;
&lt;/blockquote&gt;

&lt;p&gt;You want it to name the priority list and a concrete security line. If it shrugs, or invents rules you never wrote:&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;Check the file is in the project root (&lt;code&gt;AGENTS.md&lt;/code&gt;) or under &lt;code&gt;.cursor/rules/&lt;/code&gt; (hidden folder — easy to miss when copying).&lt;/li&gt;
&lt;li&gt;For Cursor, confirm the always-on rule has &lt;code&gt;alwaysApply: true&lt;/code&gt; in the front matter.&lt;/li&gt;
&lt;li&gt;For Claude Code, confirm &lt;code&gt;CLAUDE.md&lt;/code&gt; is at the root and that any &lt;code&gt;@AGENTS.md&lt;/code&gt; import resolves.&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;Then ask a scoped question in a file that should trigger a globbed rule, for example open &lt;code&gt;package.json&lt;/code&gt; and ask "what do my dependency rules say about adding a package?" If it cannot quote them, the glob is wrong.&lt;/p&gt;

&lt;h2&gt;
  
  
  6. A starter layout that works
&lt;/h2&gt;



&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;AGENTS.md                 # shared rules every tool can read
CLAUDE.md                 # short; @AGENTS.md + security backup
.cursor/rules/
  security.mdc            # alwaysApply: true
  testing.mdc             # globs for tests
  dependencies.mdc        # globs: **/package.json
  …stack files…           # globs for components, routes, migrations
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;You do not need every file on day one. Start with &lt;code&gt;AGENTS.md&lt;/code&gt; + one always-on security rule. Add globbed rules when the same mistake shows up twice.&lt;/p&gt;

&lt;h2&gt;
  
  
  Free starter (and a fuller kit)
&lt;/h2&gt;

&lt;p&gt;A free MIT starter with &lt;code&gt;AGENTS.md&lt;/code&gt;, &lt;code&gt;CLAUDE.md&lt;/code&gt;, a Cursor security rule and a short checklist lives here:&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;&lt;a href="https://github.com/theodorecreatesco/vibe-coding-rules-starter" rel="noopener noreferrer"&gt;github.com/theodorecreatesco/vibe-coding-rules-starter&lt;/a&gt;&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;Clone it, copy the three files into a project, run the test prompt above. Issues and PRs welcome.&lt;/p&gt;

&lt;p&gt;If you want the same rules already tuned for a stack (Next.js + Supabase, Lovable / Bolt, FastAPI, Expo and others), with native files for Cursor, Claude Code, Copilot, Windsurf, Replit, Lovable and Bolt, plus 42 copy-paste prompts and a one-command installer, I sell a &lt;a href="https://www.etsy.com/listing/4601953941" rel="noopener noreferrer"&gt;Vibe Coding Rules Kit on Etsy&lt;/a&gt;. It is &lt;strong&gt;60% off until Monday 9 Nov&lt;/strong&gt; (then back to the regular price). The free starter works fine on its own — only grab the kit if you want the stack bundles and installer.&lt;/p&gt;

&lt;p&gt;&lt;em&gt;Whatever rules you use, review AI-written code, especially anything touching auth, payments or personal data.&lt;/em&gt;&lt;/p&gt;

</description>
      <category>cursor</category>
      <category>ai</category>
      <category>coding</category>
      <category>productivity</category>
    </item>
    <item>
      <title>How to stop Cursor and Claude Code breaking your project: a free AGENTS.md starter</title>
      <dc:creator>Theodore Creates Co.</dc:creator>
      <pubDate>Sat, 31 Oct 2026 18:49:50 +0000</pubDate>
      <link>https://dev.to/theodorecreatesco/how-to-stop-cursor-and-claude-code-breaking-your-project-a-free-agentsmd-starter-2273</link>
      <guid>https://dev.to/theodorecreatesco/how-to-stop-cursor-and-claude-code-breaking-your-project-a-free-agentsmd-starter-2273</guid>
      <description>&lt;p&gt;If you build with Cursor, Claude Code, Copilot or Windsurf, you've probably had at least one of these happen:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;You asked for a small fix and the agent "tidied up" four other files.&lt;/li&gt;
&lt;li&gt;An API key ended up hard-coded in a component, or behind a &lt;code&gt;NEXT_PUBLIC_&lt;/code&gt; prefix where every visitor can read it.&lt;/li&gt;
&lt;li&gt;A new API route works perfectly, for every user, including the ones who shouldn't see that record.&lt;/li&gt;
&lt;li&gt;The agent tried the same broken fix five times, each one making things a bit worse.&lt;/li&gt;
&lt;li&gt;Tests were "fixed" by skipping them.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;None of this is the model being stupid. It's the model doing exactly what it was asked, with no idea what &lt;em&gt;you&lt;/em&gt; consider off-limits. The fix is cheap: give it a short, clear set of house rules it reads at the start of every session.&lt;/p&gt;

&lt;h2&gt;
  
  
  Where the rules go
&lt;/h2&gt;

&lt;p&gt;Most AI coding tools now read a plain Markdown instructions file from your project:&lt;/p&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Tool&lt;/th&gt;
&lt;th&gt;File it reads&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;Cursor, Copilot, Windsurf, Codex, Zed, Amp and others&lt;/td&gt;
&lt;td&gt;&lt;code&gt;AGENTS.md&lt;/code&gt;&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Cursor (project rules)&lt;/td&gt;
&lt;td&gt;&lt;code&gt;.cursor/rules/*.mdc&lt;/code&gt;&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Claude Code&lt;/td&gt;
&lt;td&gt;&lt;code&gt;CLAUDE.md&lt;/code&gt;&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;&lt;code&gt;AGENTS.md&lt;/code&gt; is the closest thing to a shared standard, so it makes sense to put the real rules there and have the other files point at it. That's the setup I'll walk through below. All the files are in a free, MIT-licensed repo: &lt;strong&gt;&lt;a href="https://github.com/theodorecreatesco/vibe-coding-rules-starter" rel="noopener noreferrer"&gt;vibe-coding-rules-starter&lt;/a&gt;&lt;/strong&gt;.&lt;/p&gt;

&lt;h2&gt;
  
  
  1. Give the agent a rule priority
&lt;/h2&gt;

&lt;p&gt;The most useful thing you can tell an agent is what to do when instructions clash. Without that, "just make it work" quietly beats "don't commit secrets". The starter opens with this:&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;## Rule priority&lt;/span&gt;

When rules clash, follow the higher one and say which you followed:
&lt;span class="p"&gt;
1.&lt;/span&gt; &lt;span class="gs"&gt;**Security guardrails**&lt;/span&gt; (next section): never broken, even if I ask.
   Explain the risk in one line and do the safe version.
&lt;span class="p"&gt;2.&lt;/span&gt; &lt;span class="gs"&gt;**The task I asked for**&lt;/span&gt;: it's my OK for the changes it clearly needs,
   within the guardrails. List them under "Needs your OK".
&lt;span class="p"&gt;3.&lt;/span&gt; &lt;span class="gs"&gt;**Stack conventions**&lt;/span&gt; (the conventions already used in this repo).
&lt;span class="p"&gt;4.&lt;/span&gt; &lt;span class="gs"&gt;**Working style.**&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;"Say which you followed" matters. It turns a silent decision into one you can see in the reply.&lt;/p&gt;

&lt;h2&gt;
  
  
  2. Make security rules specific
&lt;/h2&gt;

&lt;p&gt;"Write secure code" does nothing. Rules the agent can act on look like this:&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="p"&gt;-&lt;/span&gt; Public prefixes (&lt;span class="sb"&gt;`NEXT_PUBLIC_`&lt;/span&gt;, &lt;span class="sb"&gt;`VITE_`&lt;/span&gt;, &lt;span class="sb"&gt;`EXPO_PUBLIC_`&lt;/span&gt;, &lt;span class="sb"&gt;`PUBLIC_`&lt;/span&gt;) ship values
  to every user: never put a secret behind one. Work that needs a secret moves
  to a server or edge function.
&lt;span class="p"&gt;-&lt;/span&gt; Anything touching private data checks the caller (401 if unknown) and that they
  may touch this record (404 if missing or not theirs; 403 if they lack the role).
  Hiding a button isn't security.
&lt;span class="p"&gt;-&lt;/span&gt; Admin features need sign-in plus a role stored in the database and checked on
  the server, never a client flag or hard-coded emails or IDs.
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;The second rule is the one that prevents the most common hole in AI-built apps: an endpoint that checks you're logged in but not that the record is &lt;em&gt;yours&lt;/em&gt;. Change the ID in the URL and you're reading someone else's data.&lt;/p&gt;

&lt;p&gt;It also helps to list the things the agent must never do on its own:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight markdown"&gt;&lt;code&gt;Never, in any mode, unless I ask for that exact action
&lt;span class="p"&gt;-&lt;/span&gt; Destructive commands (&lt;span class="sb"&gt;`rm -rf`&lt;/span&gt;, &lt;span class="sb"&gt;`git reset --hard`&lt;/span&gt;, &lt;span class="sb"&gt;`git push --force`&lt;/span&gt;, dropping
  tables, database resets), changing a remote or production database, or disabling,
  skipping or loosening tests, lint rules or type checks to get a pass.
&lt;span class="p"&gt;-&lt;/span&gt; Installing new dependencies without my OK. Check a package's exact name before
  installing (typo-squats are common).
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;h2&gt;
  
  
  3. Keep it in scope and stop the spiral
&lt;/h2&gt;

&lt;p&gt;Most "the AI broke my project" stories are scope problems, not security ones. These lines do most of the work:&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="p"&gt;-&lt;/span&gt; Stay in scope: don't rename, reformat or tidy unrelated code.
&lt;span class="p"&gt;-&lt;/span&gt; Make the smallest change; never rewrite a working file wholesale.
&lt;span class="p"&gt;-&lt;/span&gt; Never invent APIs, packages or file paths. Check the installed version's docs.
&lt;span class="p"&gt;-&lt;/span&gt; Debugging: reproduce, read the full error, state the likely cause, make the
  smallest fix. After two failed attempts, stop and summarise the next two options.
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;The two-attempt limit is what stops the debugging spiral. Instead of a sixth guess, you get a summary and a choice.&lt;/p&gt;

&lt;p&gt;For anything bigger than two files, the agent writes a short plan first and waits for your OK. If it's running as a background agent and can't ask, it puts the plan at the top of its summary instead of stalling.&lt;/p&gt;

&lt;h2&gt;
  
  
  4. Ask for a report you can check
&lt;/h2&gt;

&lt;p&gt;Finally, tell the agent what "done" looks like:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight markdown"&gt;&lt;code&gt;Reply with: 1) what changed, 2) the files changed, 3) the commands you ran and
their results, 4) how I can test it, 5) "Needs your OK": anything I should
confirm, plus anything risky or unfinished.
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Point 3 is the important one. "I ran &lt;code&gt;npm test&lt;/code&gt;, 41 passed, lint couldn't run becauseâ€¦" is something you can trust. "Everything should work now" isn't.&lt;/p&gt;

&lt;h2&gt;
  
  
  Making Cursor and Claude Code load it every time
&lt;/h2&gt;

&lt;p&gt;&lt;strong&gt;Cursor.&lt;/strong&gt; &lt;code&gt;AGENTS.md&lt;/code&gt; is picked up automatically, but a project rule with &lt;code&gt;alwaysApply: true&lt;/code&gt; makes sure the security section is in every chat, even if &lt;code&gt;AGENTS.md&lt;/code&gt; gets cut short:&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="nn"&gt;---&lt;/span&gt;
&lt;span class="na"&gt;description&lt;/span&gt;&lt;span class="pi"&gt;:&lt;/span&gt; &lt;span class="s"&gt;Rule priority and security guardrails that apply to every task&lt;/span&gt;
&lt;span class="na"&gt;alwaysApply&lt;/span&gt;&lt;span class="pi"&gt;:&lt;/span&gt; &lt;span class="kc"&gt;true&lt;/span&gt;
&lt;span class="nn"&gt;---&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;&lt;strong&gt;Claude Code.&lt;/strong&gt; &lt;code&gt;CLAUDE.md&lt;/code&gt; can import another file with &lt;code&gt;@&lt;/code&gt;, so it stays short and repeats only the security essentials as a backup:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight markdown"&gt;&lt;code&gt;@AGENTS.md

AGENTS.md (imported above) holds the full project rules. This file repeats the
essentials so they still apply if AGENTS.md is missing or cut short; if they
ever differ, the security rules here win.
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;h2&gt;
  
  
  30-second setup
&lt;/h2&gt;

&lt;p&gt;From the root of your project:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;git clone &lt;span class="nt"&gt;--depth&lt;/span&gt; 1 https://github.com/theodorecreatesco/vibe-coding-rules-starter.git /tmp/vcrs
&lt;span class="nb"&gt;cp&lt;/span&gt; /tmp/vcrs/AGENTS.md /tmp/vcrs/CLAUDE.md &lt;span class="nb"&gt;.&lt;/span&gt;
&lt;span class="nb"&gt;mkdir&lt;/span&gt; &lt;span class="nt"&gt;-p&lt;/span&gt; .cursor/rules &lt;span class="o"&gt;&amp;amp;&amp;amp;&lt;/span&gt; &lt;span class="nb"&gt;cp&lt;/span&gt; /tmp/vcrs/.cursor/rules/security.mdc .cursor/rules/
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Already have an &lt;code&gt;AGENTS.md&lt;/code&gt; or &lt;code&gt;CLAUDE.md&lt;/code&gt;? Don't overwrite it. Paste the block between the &lt;code&gt;BEGIN&lt;/code&gt; and &lt;code&gt;END&lt;/code&gt; markers at the top of yours.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Check it worked:&lt;/strong&gt; start a new chat and ask &lt;em&gt;"What are the top rules in this project?"&lt;/em&gt; It should mention the rule priority and the security guardrails. If it doesn't, the file isn't being loaded. Check it's in the project root, and for Cursor, that the &lt;code&gt;.cursor&lt;/code&gt; folder (hidden by default) was copied.&lt;/p&gt;

&lt;h2&gt;
  
  
  Before you share your app
&lt;/h2&gt;

&lt;p&gt;Rules reduce mistakes, they don't guarantee anything. The repo also includes a 10-point checklist to run before you share a link. A few of the checks:&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="p"&gt;-&lt;/span&gt; [ ] &lt;span class="sb"&gt;`.env`&lt;/span&gt; and &lt;span class="sb"&gt;`.env.*`&lt;/span&gt; are listed in &lt;span class="sb"&gt;`.gitignore`&lt;/span&gt; (and &lt;span class="sb"&gt;`.env.example`&lt;/span&gt; holds names only).
      Check: &lt;span class="sb"&gt;`git check-ignore -q .env.local`&lt;/span&gt; should exit with code 0.
&lt;span class="p"&gt;-&lt;/span&gt; [ ] Every API route, server action and edge function checks the user is logged in
&lt;span class="p"&gt;-&lt;/span&gt; [ ] Every one also checks the user owns (or may access) the record
&lt;span class="p"&gt;-&lt;/span&gt; [ ] Spending limits or alerts are set on paid APIs (OpenAI, Anthropic, etc.) and hosting
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;There's also a copy-paste "secret scan" prompt that gets your agent to check the repo (and git history, if it can) for keys, without printing the values.&lt;/p&gt;

&lt;h2&gt;
  
  
  Wrapping up
&lt;/h2&gt;

&lt;p&gt;You don't need a huge rules file. A clear priority order, specific security rules, a scope limit and a "show me what you ran" report fix most of the damage AI coding tools do. Grab the files from &lt;strong&gt;&lt;a href="https://github.com/theodorecreatesco/vibe-coding-rules-starter" rel="noopener noreferrer"&gt;vibe-coding-rules-starter&lt;/a&gt;&lt;/strong&gt;, and if you improve on them, issues and PRs are welcome.&lt;/p&gt;

&lt;p&gt;If you want versions tuned for a specific stack (Next.js + Supabase, Lovable, Bolt, FastAPI, Expo and others) with a one-command installer, I also sell a &lt;a href="https://www.etsy.com/uk/listing/4601953941" rel="noopener noreferrer"&gt;full Vibe Coding Rules Kit&lt;/a&gt;. The free starter works fine on its own.&lt;/p&gt;

&lt;p&gt;&lt;em&gt;Whatever rules you use, review AI-written code, especially anything touching auth, payments or personal data.&lt;/em&gt;&lt;/p&gt;

</description>
      <category>ai</category>
      <category>cursor</category>
      <category>claude</category>
      <category>security</category>
    </item>
  </channel>
</rss>
