agentleFS
SamuelBostic29SamuelBostic29/claude-skills/CLAUDE.md

claude-skills

CLAUDE.md1 starsChanged 4 months agoMITView on GitHub

A curated, public collection of portable Claude Code skills — the sharable subset of a larger personal toolkit. Everything here is portfolio-grade and org/machine-agnostic. Treat every change as public-facing. - One job. A single, clear purpose — if the description needs "and", split it. - Decision rules, not vibes. Hard caps, explicit "do X / NEVER Y", stop conditions. - Convergent. Skills reduce churn; output is review-ready, never make-work. - Portable. No org names, hosts, real paths, or machine specifics.…

The file

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

7 rules. About 0 are common, 2 are distinctive, and 5 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 5

    Very specific

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

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

    Portable: No org names, hosts, real paths, or machine specifics.

    line 12
  2. Very specific:

    Evals mandatory: Every skill ships evals/README.md — 2–3 representative cases, at least one negative ("should not fire / should ask …

    line 15
  3. Distinctive:

    Add or update the skill's row in the README catalog, in the right group.

    line 26
  4. Very specific:

    Run python scripts/validate-skills.py — it must exit 0 (CI runs it on every PR).

    line 27
  5. Very specific:

    Never commit docs/plans/ (local planning docs), tokens, or local config.

    line 33
  6. Distinctive:

    Never introduce org/employer specifics — this repo is public.

    line 34
  7. Very specific:

    Never ship a skill without evals, or with surviving <ANGLE_BRACKET> placeholders (template/ is the one intentional exception).

    line 35

≈ “Common” is an estimate: at least 1 in 5 of the 27,641 CLAUDE.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
CLAUDE.md

CLAUDE.md — claude-skills

A curated, public collection of portable Claude Code skills — the sharable subset of a larger personal toolkit. Everything here is portfolio-grade and org/machine-agnostic. Treat every change as public-facing.

Quality bar (every skill, no exceptions)

  • One job. A single, clear purpose — if the description needs "and", split it.
  • Decision rules, not vibes. Hard caps, explicit "do X / NEVER Y", stop conditions.
  • Convergent. Skills reduce churn; output is review-ready, never make-work.
  • Portable. No org names, hosts, real paths, or machine specifics. Anything project-specific is a <PLACEHOLDER> or a Configuration-table entry the adopter fills in. Stack-specific skills (e.g. .NET) are labeled in the README.
  • Evals mandatory. Every skill ships evals/README.md — 2–3 representative cases, at least one negative ("should not fire / should ask instead").
  • Frontmatter: name equals the folder name; description leads with the trigger; allowed-tools is the minimal set (scope Bash to subcommands where possible).

Workflow for adding or changing a skill

  1. Brief it as a GitHub issue — this repo's issues are skill briefs.
  2. Branch skill/<name> off main. Start from the template (cp -r template skills/<name>) or run the skill-builder skill.
  3. Add or update the skill's row in the README catalog, in the right group.
  4. Run python scripts/validate-skills.py — it must exit 0 (CI runs it on every PR).
  5. Open a draft PR (Closes #<issue>); merge only after review.

Never

  • Never commit docs/plans/ (local planning docs), tokens, or local config.
  • Never introduce org/employer specifics — this repo is public.
  • Never ship a skill without evals, or with surviving <ANGLE_BRACKET> placeholders (template/ is the one intentional exception).

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.