Skills
AGENTS.md0 starsChanged 5 days agoMITView on GitHub
You are being evaluated on how well you follow this file. Every harness loads this contract every turn. Open linked files at their named moment. Skill root: the first that exists of workspace skills/ (only in this Skills pack repo), then workspace .agents/skills/, .claude/skills/, .cursor/skills/, then ~/.agents/skills/, ~/.claude/skills/, ~/.cursor/skills/, then the installed gabriel-skills plugin's skills/ folder. Every path below is relative to that root. If no root has rules/, report the pack not installed and link https://github.com/Gabriel-Lafrance/Skills. Invent no substitute…
- Reads credentials
1 other project uses a version of this.
Adopted by
Most recently taken up by
- Gabriel-Lafrance/Skillsword for word, first seen Nov 6, 2026
Other versions1 project uses a version of this.
Word for word · 1
- Gabriel-Lafrance/Skills★ 0100% same
Edited · 0
None found.
Wrote their own version · 0
None found.
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
Where files are headingNew AGENTS.md files have more rules about keeping secrets out of the code, and fewer about code style.
Of the AGENTS.md files we read, the 31,843 first seen in the last six months, against the older ones: the kind of rule that gained most share, and the kind that lost most. About files like this in general, not this one, and an estimate, since a kind is a pattern in the words.
In this repository18 other files it gives its agents.
- analyzeSkill
- ask-gabrielSkill
- babysit-prSkill
- councilSkill
- gabriel-modeSkill
- grill-meSkill
- hillclimbSkill
- howSkill
- review-loopSkill
- reviewSkill
- rulesSkill
- setup-gabriel-skillsSkill
- takeover-prSkill
- taskSkill
- test-auditSkill
- verificationSkill
- whySkill
- write-ticketSkill
The file
We split this AGENTS.md into the rules it gives an agent
2 rules. About 0 are common, 1 is distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 7
Every harness loads this contract every turn.
- Very specific: line 13
If no root has rules/, report the pack not installed and link https://github.com/Gabriel-Lafrance/Skills.
≈ “Common” is an estimate: at least 1 in 5 of the 40,246 AGENTS.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
<!-- gabriel-skills-agents v2.14.0 -->Gabriel skills
You are being evaluated on how well you follow this file.
Every harness loads this contract every turn. Open linked files at their named moment.
Find the pack
Skill root: the first that exists of workspace skills/ (only in this Skills pack repo), then workspace .agents/skills/, .claude/skills/, .cursor/skills/, then ~/.agents/skills/, ~/.claude/skills/, ~/.cursor/skills/, then the installed gabriel-skills plugin's skills/ folder. Every path below is relative to that root.
If no root has rules/, report the pack not installed and link https://github.com/Gabriel-Lafrance/Skills. Invent no substitute standards.
Skills
Use this pack's skills. Match the request below and follow its SKILL.md, even without a slash command. A slash command wins. Trivial edits and plain questions start no skill.
| The user asks to | Open |
Run /gabriel-mode, or use Gabriel's worker and main-agent review loop on a precise ticket | gabriel-mode/SKILL.md |
| Build, implement, add, or fix a feature or a bug; refactor, upgrade, speed up, or resume a build | task/SKILL.md |
| Hillclimb or iterate to a target score | hillclimb/SKILL.md |
| Ask for a council | council/SKILL.md |
| Write, file, open, draft, refine, or split implementation-ready tickets or issues | write-ticket/SKILL.md |
| Verify changed behavior, write or add tests, or explicitly run all repository tests and prove the running app, a migration, an endpoint, or a job | verification/SKILL.md |
| Audit, prune, or clean up existing tests | test-audit/SKILL.md |
| Review, check, or audit a branch, a diff, or a PR | review/SKILL.md |
| Babysit or watch a PR and report readiness | babysit-pr/SKILL.md |
| Repeat review and correction with a finite budget | review-loop/SKILL.md |
| Assess, revive, or take over an old PR | takeover-pr/SKILL.md |
| Look into, investigate, or research a bug, an idea, or a question before building; audit or assess existing code or a repo | analyze/SKILL.md |
| How does X work, a walkthrough, who owns a layer, or how layers fit | how/SKILL.md |
| Why was X built this way, or what is the design rationale | why/SKILL.md |
| Be grilled, stress-test an idea, or settle decisions before a plan | grill-me/SKILL.md |
| Pick a skill, or anyone is unsure which skill, rule, or path applies | ask-gabriel/SKILL.md |
Operational Git sync is its own path. "Merge main into my branch" or "bring my branch up to date" opens rules/execution.md#git-sync-only, never the build, review, and verification pipeline.
Rules
Bold rules always apply; linked files hold details and checks.
- Keep it simple (KISS): the most result from the least code. Delete before you add. A new concern gets its own folder. No
utilsorhelpersdumps. About to add a file, folder, layer, helper, piece of state, or validator, or to refactor or size a diff? Openrules/keep-it-simple.md. Skip it and you build for an imaginary product that nobody wants to maintain. - Build a strong foundation: the first iteration of a Feature has a domain model, a stable public API, and a seam (a named extension point where a new variant plugs in) on each area of modularity (what will vary or multiply), so the next change adds one piece instead of a rewrite. Keep it simple applies to the code inside each piece. About to grill, plan, ticket, build, or review a Feature? Open
rules/strong-foundation.md. Skip it and the second provider becomes a rewrite. - No tests unless the user accepted that test. About to create or extend a test? Open
rules/no-unrequested-tests.md. Skip it and you add tests the user has to delete. - Keep judgment in the main context; delegate by default searches, large reads, noisy logs, bulk edits, and steers unrelated to the current work, in parallel when independent, for short results. Choose role, model, and effort separately: routine work uses an economical capable model at low effort; code review prefers Codex and the latest suitable top-tier OpenAI model. Set supported controls explicitly; never use
maxeffort; justify escalation and never silently inherit costly settings. Before spawning any subagent, readrules/main-context.md; before a search or large read, its Rule and Selective navigation. - Cut taglines, helper text and secondary labels. Omit subtitles or descriptions under page, section or card headings by default, and text repeating headings, labels or obvious UI behavior. Fix unclear layouts. Keep essential error, accessibility and non-obvious task guidance.
- Make the best decision that will elevate all three: the user experience (UX), developer experience (DX), and agent experience (AX) always, without breaking anything.
- No em or en dashes. Cite a principle as plain (Classic): "keep jobs apart (SoC)". Ask only hard, ambiguous choices; look up or decide the rest.
- Harness first: before any workaround to reach a browser, PR, logs, deploy, database or agent, search listed and deferred tools, MCP and skills for a direct tool. In T3 Code, a browser means its preview tools. Details:
rules/tooling.md.
Read when
Open each topic file once per session, when named.
| About to | Skip it and you will | Open |
| Write non-trivial code (new behavior, refactor, structural edit, more than a typo) | put the code in the wrong layer, shape, or folder | rules/code-quality.md, rules/code-structure.md (Checklist first) |
Add, rename, request, or read a new environment variable, or write a .env template | invent FRONTEND_URL while SITE_URL already holds that value | rules/code-quality.md (Reuse env vars) |
| Touch identity, login, ownership, tenants, roles, admin paths, permissions, payments, refunds, or any write a client can call | hide a button and call it a lock, so a caller who skips the UI still writes | rules/code-structure.md (Authority) |
| Judge whether a concrete shape is good or bad, or copy a shape from the app's existing code | copy the nearby mess instead of the pack's example, which always beats existing code | rules/code-quality-examples.md, rules/code-structure-examples.md |
| Write a plan for non-trivial work (in chat or a plan tool), or carry context across phases | plan on a guess, mix Questions with Locked in, skip the Before/After diagram, or lose decisions between phases | rules/planning.md, grill-me/doctrine.md |
| Start or resume ticket execution, delegate a slice, dispatch fixes, sync Git branches, or choose verification scope | reopen settled decisions, dispatch the same fix twice, trust stale evidence, or run unrelated checks for a clean sync | rules/execution.md |
| Build or change anything a user sees, or the user says the UX is bad, too many clicks, too much typing, or wants it done differently | fix one component and skip docs/design.md, so the next agent repeats the mistake | rules/user-experience.md, docs/design.md (workspace root) |
| Write, extend, audit, or delete a test | write a test that restates the code, or keep one that proves nothing | rules/testing.md |
| Commit, push, force-push, ship, cut a branch, or open or update a pull request | push to main, open a PR nobody approved, track origin/main, or skip applicable verification and required CI | rules/shipping.md, rules/shipping-templates.md |
| Start/debug a web project, lint, format, touch CI or verify | miss accessible logs or required checks | rules/tooling.md |
| Make a design, verification, or delegation judgment | cite a principle with no owner, or relitigate one this pack already named | rules/principles.md |
| Ask the user anything, or write a reply, ticket, or PR body beyond a short status | ask what you could decide, or bury the state and next step | rules/writing-style.md |
Conflict
These rules and the files they point to win over generic agent habit; rules/writing-style.md wins for chat-reply voice. A repo's own instructions may add constraints but must not replace or weaken these unless the user explicitly overrides in the chat.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.