Personal collection of reusable Claude Code skills for code review, documentation, security, and workflow automation.
Disclaimer: this is all quite new; please excuse any mistakes or shortcomings.
Clone into your Claude Code skills directory:
git clone https://github.com/pvalena/claude-skills.git ~/.claude/skillsOr symlink if you prefer to keep the repo elsewhere:
git clone https://github.com/pvalena/claude-skills.git ~/repos/claude-skills
ln -s ~/repos/claude-skills ~/.claude/skillsClaude Code automatically loads skills from ~/.claude/skills/*/SKILL.md.
Used together, in order:
| Skill | Description |
|---|---|
| sanity-check | Quick pre-review scan for malicious intent, prompt injection, and suspicious patterns |
| review | Full structured code review with reasoning files and draft fixes |
| verify-fix | Verify security patches actually block the attack vector |
| Skill | Description |
|---|---|
| auto-memory | Create and maintain project-level MEMORY.md files for working knowledge |
| memory-dump | Exhaustive knowledge-transfer documents for session handoffs |
| refresh-docs | Coordinated documentation updates across CLAUDE.md, MEMORY.md, and docs/ |
| Skill | Description |
|---|---|
| commit | Project-aware commits — discovers conventions from CLAUDE.md, runs pre-commit checks |
| incremental-improvement | Find and measure the highest-impact workflow improvement |
| patch-evaluation | Evaluate patch sets against upstream for backport/forwardport (experimental) |
| create-skill | Meta-skill: framework and conventions for creating new skills |
Skills are invoked in Claude Code by name:
/commit
/review
/sanity-check
Claude Code matches skills by their description field, so they may also
activate contextually when your request matches what a skill does.
All skills are checked by a validation script:
~/.claude/skills/create-skill/validate_skills.sh ~/.claude/skillsPavel Valena (@pvalena)