agentleFS
albertograndealbertogrande/readme-skill/AGENTS.md

readme-skill

AGENTS.md2 starsChanged 4 months agoMITView on GitHub

Guidance for coding agents working on readme-skill itself. Human-facing overview is in README.md; this file holds only what you can't infer from reading the tree. (This repo practices what it documents — see ai-readers.md.) The knowledge base must live inside skills/readme/ (references/ and templates/). Do not move it to a top-level guide/ or docs/. When the plugin is installed, only the plugin directory is copied to a cache; a skill cannot read files outside it. skills/readme/SKILL.md references its siblings via…

  • Commits and pushes
  • 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 32,671 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 repository4 other files it gives its agents.

The file

We split this AGENTS.md into the rules it gives an agent

5 rules. About 0 are common, 1 is distinctive, and 4 are 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.

  • 4

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Not sourced, so it is exempt from the Last verified/Sources rule below.

    line 26
  2. Very specific:

    Every file in skills/readme/references/ and templates/:

    line 32
  3. Very specific:

    Check every KB Sources: key resolves in sources.md.

    line 51
  4. Very specific:

    Follow those protocols rather than editing the KB freehand: diff findings against current content, change only contradictions/additions, …

    line 59
  5. Distinctive:

    Keep edits diff-legible: no repo-wide rewording sweeps that bury real changes.

    line 81

≈ “Common” is an estimate: at least 1 in 5 of the 41,581 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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
AGENTS.md

AGENTS.md

Guidance for coding agents working on readme-skill itself. Human-facing overview is in README.md; this file holds only what you can't infer from reading the tree. (This repo practices what it documents — see ai-readers.md.)

The one architecture rule that will bite you

The knowledge base must live inside skills/readme/ (references/ and templates/). Do not move it to a top-level guide/ or docs/. When the plugin is installed, only the plugin directory is copied to a cache; a skill cannot read files outside it. skills/readme/SKILL.md references its siblings via ${CLAUDE_SKILL_DIR} — that path resolves to the installed copy, so the KB has to travel with the skill. The README is the human table of contents that links into it; that is the only intended duplication of structure.

What ships vs what doesn't

  • skills/readme/ → shipped in the plugin (the consumer-facing skill).
  • .claude/skills/refresh-research/ → not shipped. It lives outside skills/ on purpose so consumers never receive the maintainer tooling. Do not relocate it into skills/.
  • skills/readme/preferences/ → shipped. The style-profile subsystem (concept, template, example) — the subjective layer over the cited standard. Not sourced, so it is exempt from the Last verified/Sources rule below.
  • research/log/ → provenance, not shipped logic. Never edit past log entries; append new ones.

Knowledge-base file conventions

Every file in skills/readme/references/ and templates/:

  • Starts with > Last verified: YYYY-MM-DD. Only bump this date when the file's sources were actually re-fetched — editing prose is not verifying.
  • Ends with a Sources: key1, key2 line. Every key MUST exist as a row in sources.md. If you cite something new, add its row to sources.md first.
  • SKILL.md is routing only — it points at reference files; never inline KB content into it. Keep it under ~150 lines.

"Tests" (there is no build system)

This is markdown + skills, so verification is:

  1. claude plugin validate . — must pass clean before any commit that touches manifests or skill structure.
  2. Dogfood: run the readme skill's review mode on a changed template/reference using a fresh subagent, and (for generate changes) the writer/reviewer split — the pattern recorded in CHANGELOG under 0.1.0 "Verified".
  3. Check every KB Sources: key resolves in sources.md.

Refresh workflow

Content changes to the KB normally come from /refresh-research ([full|quick|source-check]), which runs research-protocol.md and update-protocol.md. Follow those protocols rather than editing the KB freehand: diff findings against current content, change only contradictions/additions, record every run in research/log/ and CHANGELOG.md.

Release process

Two tags per release, same commit:

  • vX.Y.Z — the human GitHub Release (drives the release badge).
  • readme-skill--vX.Y.Z — created by claude plugin tag --push, which validates that plugin.json and the marketplace.json entry agree. Run the --dry-run first.

Bump version in .claude-plugin/plugin.json per update-protocol.md: patch for KB contradictions/additions, minor for a new template or reference file.

Commit / PR conventions

  • Conventional, imperative subject lines. Prefix refresh commits with research:.
  • A claim only enters the KB with a citable source or a reproducible observation. No "best practice" assertions without provenance.
  • Keep edits diff-legible: no repo-wide rewording sweeps that bury real changes.

Environment note

Pushing uses HTTPS + gh auth git-credential (configured in this repo's git config). The machine's SSH key maps to a different GitHub identity than the gh login, so plain git push over SSH can fail — the HTTPS helper is intentional.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.