agentleFS
luizomfluizomf/omskills/AGENTS.md

omskills

AGENTS.md3 starsChanged 2 months agoMITView on GitHub

This repository contains agent skills, prompts, docs, and helper scripts—not an application. This file is canonical; CLAUDE.md is a compatibility pointer. Use English for repository artifacts and the user's language in chat. Read the context relevant to the request, not every workflow document. Workflows provided by skills are not mandatory steps for ordinary repository maintenance. Adopt upstream ideas deliberately; do not synchronize wholesale with the source project. Keep this file short and repository-specific, not a workflow manual. - Carry the…

The file

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

10 rules. About 0 are common, 9 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈9

    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
  • could be automatic: a setting could enforce it
  1. Distinctive:

    Use English for repository artifacts and the user's language in chat.

    line 5
  2. Distinctive:

    Read the context relevant to the request, not every workflow document.

    line 9
  3. Distinctive:

    Keep going when the next step is clear and in scope.

    line 16
  4. Distinctive:

    Use disposable notes or transport files when useful under Working files.

    line 23
  5. Distinctive:

    Prefer the smallest clear, maintainable solution.

    line 26
  6. Very specific:

    Prefer portable Linux/macOS paths and scripts.

    line 31
  7. Distinctive:

    Do not force-push, rewrite history, publish releases, or change repository visibility without explicit authorization.

    line 33
  8. Distinctive:

    When you encounter a concrete problem outside the current task—such as confusing navigation, conflicting instructions, broken tooling, or …

    line 35
  9. Distinctive:

    Use relative Markdown links for cross-skill loading.

    line 59
  10. Distinctive:

    Run checks relevant to the change:

    line 69

≈ “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.

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

Working on omskills

This repository contains agent skills, prompts, docs, and helper scripts—not an application. This file is canonical; CLAUDE.md is a compatibility pointer. Use English for repository artifacts and the user's language in chat.

Working style

Read the context relevant to the request, not every workflow document. Workflows provided by skills are not mandatory steps for ordinary repository maintenance. Adopt upstream ideas deliberately; do not synchronize wholesale with the source project. Keep this file short and repository-specific, not a workflow manual.

  • Carry the requested work through verification, conventional commit, and push to origin, unless the user asks otherwise. No branch or PR ceremony needed.
  • Keep going when the next step is clear and in scope. Ask only when missing input blocks progress or an action is destructive or hard to reverse.
  • Translate informal maintainer intent into the affected skills and governing docs; exact skill names are not required. Before delivery, check connected workflows for conflicting or ambiguous instructions introduced by the change.
  • Inspect Git status and relevant files before editing. Preserve unrelated work; never commit secrets, private notes, local settings, or scratch artifacts.
  • Use disposable notes or transport files when useful under Working files; keep lasting outcomes in the authorized durable destination, not scratch.
  • Prefer the smallest clear, maintainable solution. Avoid speculative features, dependencies, abstractions, and process. Fix relevant outdated docs too.
  • Review the diff and run checks relevant to the change. Prefer test-first for behavior changes and regression tests for bugs; test observable behavior, not incidental wording. Documentation edits do not need invented tests.
  • Prefer portable Linux/macOS paths and scripts. Put new worktrees under ~/sannux-data/worktrees/<repo>/<worktree_name> if needed.
  • Do not force-push, rewrite history, publish releases, or change repository visibility without explicit authorization.
  • When you encounter a concrete problem outside the current task—such as confusing navigation, conflicting instructions, broken tooling, or misleading docs—record it in the configured issue tracker for later investigation rather than ignore it silently. Check for an existing issue first, include observed evidence and impact, distinguish observations from hypotheses, and omit sensitive data.
  • Finish with a concise summary, verification results, and any remaining issues.

Where things live

  • skills/engineering/ and skills/productivity/: skills and bucket catalogs.
  • README.md: active catalog; .codex-plugin/plugin.json: active distribution; .claude-plugin/plugin.json: matching entries in the same order.
  • CONTEXT.md and docs/adr/: domain vocabulary and durable design decisions.
  • docs/agents/: tracker and domain-document configuration when needed.

Skill catalog and installation

  • Active skills belong in the root README, bucket README, and both manifests. Optional skills belong only in their bucket README. Link entries to SKILL.md.
  • New skills default to disable-model-invocation: true; making them permanently discoverable needs maintainer approval. USER_ONLY_ACTIVE_SKILLS in scripts/check-catalog.py lists active skills that remain user-invoked only.
  • Rename the folder, frontmatter name, catalog entries, and references together. Preserve existing behavior unless the request changes it.
  • Use relative Markdown links for cross-skill loading. Resolve the referring file's symlink first, then resolve links from its physical directory—not the workspace or flat installation directory.
  • Install only what was requested, using relative symlinks from the destination's physical directory. For the full active catalog, use scripts/link-skills.sh with OMSKILLS_DEST for the chosen destination. Verify links with readlink and a readable target SKILL.md.

Checks

Run checks relevant to the change:

  • Catalog, manifest, or installer changes: ./scripts/check-catalog.py and ./tests/test-link-skills.sh.
  • Skill file references or pointer-resolution changes: ./scripts/check-catalog.py and python3 -B tests/test-skill-pointers.py.
  • Architecture-report HTML scaffold changes: python3 -B tests/test-html-report.py.

For manifest or installation changes, also run ./scripts/link-skills.sh --check when the managed destination exists. No repository-wide formatter, linter, or typechecker is configured.

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.