claude-skills
CLAUDE.md0 starsChanged 8 months agoMITView on GitHub
Monorepo for independently publishable PRPM skill packages. Each skill lives under skills/<name>/ with its own prpm.json, agents, scripts, and supporting content. Before committing, all gates must pass: CI will: 1. Update version in skills/review-code/prpm.json 2. Publish to PRPM 3. Create GitHub Release 4. Push version bump commit to main If release fails with auth errors, refresh the PRPM token:
- Commits and pushes
In this repository2 other files it gives its agents.
- agent-evalSkill
- review-codeSkill
The file
Show its rules and text
We split this CLAUDE.md into the rules it gives an agent
14 rules. About 0 are common, 7 are distinctive, and 7 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈7
Distinctive · estimate
This author's own choices. Worth a look.
7
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 35
Use ~/.claude/skills/dakaneye-<skill-name>/ in agent/rule files (runtime PRPM install path)
- Very specific: line 36
Use relative paths within a skill directory for file references in prpm.json
- Distinctive: line 42
Never manually edit the version field
- Very specific: line 45
All scripts use set -Eeuo pipefail
- Distinctive: line 51
Before committing, all gates must pass:
- Distinctive: line 78
All quality gates pass
- Distinctive: line 79
No hardcoded personal paths or repos
- Very specific: line 90
Update version in skills/review-code/prpm.json
- Distinctive: line 97
If release fails with auth errors, refresh the PRPM token:
- Very specific: line 106
Create skills/<name>/SKILL.md with frontmatter (name, description with trigger phrases)
- Very specific: line 107
Create skills/<name>/prpm.json with package definition
- Very specific: line 108
Create skills/<name>/evals/evals.json and trigger-evals.json
- Distinctive: line 109
Add any agents, scripts, rules, concepts, patterns under the skill directory
- Distinctive: line 111
Update README.md skill listing
≈ “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.
Files
Claude Development Guidelines for claude-skills
Project Overview
Monorepo for independently publishable PRPM skill packages. Each skill lives under skills/<name>/ with its own prpm.json, agents, scripts, and supporting content.
Structure
claude-skills/
├── skills/
│ └── <skill-name>/
│ ├── SKILL.md # Skill entry point (no version — PRPM forbids it)
│ ├── prpm.json # PRPM package definition
│ ├── evals/ # Reasoning and trigger evals
│ ├── agents/ # Specialized subagents
│ ├── scripts/ # Helper scripts
│ ├── rules/ # Language quality checklists
│ ├── concepts/ # Deep dive documentation
│ └── patterns/ # Design pattern references
├── tests/ # Linter and eval validator (scan all skills)
├── .github/workflows/ # CI + per-skill release
├── CLAUDE.md
├── README.md
├── CONTRIBUTING.md
└── LICENSEKey Patterns
Skills Are Self-Contained
Each skill directory contains everything it needs. No shared code between skills. This keeps skills independently publishable.
Path References
- Use
~/.claude/skills/dakaneye-<skill-name>/in agent/rule files (runtime PRPM install path) - Use relative paths within a skill directory for file references in
prpm.json - Scripts auto-detect repo from git remote (no hardcoded defaults)
Versioning
- Tags drive releases:
<skill-name>/v<semver>(e.g.,review-code/v1.0.5) - Release workflow auto-bumps version in
prpm.json(PRPM forbids version in SKILL.md frontmatter) - Never manually edit the version field
Scripts
- All scripts use
set -Eeuo pipefail - Auto-detect repo from git remote when possible
- Portable bash (no bashisms that break on different systems)
Quality Gates
Before committing, all gates must pass:
# 1. Validate all prpm.json files
for f in skills/*/prpm.json; do jq empty "$f"; done
# 2. Each skill has a SKILL.md
for d in skills/*/; do test -f "${d}SKILL.md"; done
# 3. Validate scripts are executable
for f in skills/*/scripts/*.sh; do test -x "$f" || echo "Not executable: $f"; done
# 4. Shellcheck scripts
shellcheck -x -S error skills/*/scripts/*.sh
# 5. Skill linter
python3 tests/skill-linter.py
# 6. Eval validator
python3 tests/eval-validator.py
# 7. Check for machine-specific references
! grep -r '/Users/' --include='*.md' --include='*.sh' --exclude='CLAUDE.md' .Pre-Push Checklist
- All quality gates pass
- No hardcoded personal paths or repos
Releasing
# Tag triggers the release — version is auto-bumped by CI
git tag -a review-code/v1.0.5 -m "review-code v1.0.5"
git push origin review-code/v1.0.5CI will:
- Update version in
skills/review-code/prpm.json - Publish to PRPM
- Create GitHub Release
- Push version bump commit to main
Token Expiry
If release fails with auth errors, refresh the PRPM token:
prpm login
gh secret set PRPM_TOKEN --repo dakaneye/claude-skills < <(jq -r '.token' ~/.prpmrc)Adding a New Skill
- Create
skills/<name>/SKILL.mdwith frontmatter (name, description with trigger phrases) - Create
skills/<name>/prpm.jsonwith package definition - Create
skills/<name>/evals/evals.jsonandtrigger-evals.json - Add any agents, scripts, rules, concepts, patterns under the skill directory
- Verify:
python3 tests/skill-linter.py && python3 tests/eval-validator.py - Update
README.mdskill listing - Tag to release:
git tag -a <name>/v1.0.0 -m "<name> v1.0.0"
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.

