Claude Code Custom Skills

Slash commands via ~/.claude/commands/ and auto-discovered skills via ~/.claude/skills/ (both symlinked from dotfiles/claude/).

Directory Structure

dotfiles/claude/
├── commands/          # User-invoked slash commands (16)
├── skills/            # Auto-discovered skills (6)
│   ├── security-audit/SKILL.md
│   ├── error-trace/SKILL.md
│   ├── data-validate/SKILL.md
│   ├── dep-audit/SKILL.md
│   ├── pipeline-check/SKILL.md
│   └── cross-dep/SKILL.md
├── CLAUDE.md
└── README.md

Commands (user-invoked)

Command	Mode	Variables
/prompt-review	Prompt quality review	prompt_files_or_text, target_model
/config-audit	YAML/env config mapping	config_files_or_packages
/stats-reconcile	Pipeline count reconciliation	pipeline_or_query, discrepancy
/uv-sync	UV workspace dependency alignment	workspace_root
/cover-letter	Job application materials	job_description, resume_or_background, company_name, tone
/uv-init	Project scaffolding	project_name, description, cli_command
/prompt-build	Draft and refine prompts	goal_or_use_case, target_model, sample_input
/plan	Plan-first development	task
/review	Code review	diff_or_pr
/fix	Fix & diagnose	(see file)
/implement	Implement mode	(see file)
/improve	Improve mode	(see file)
/iterate	Iterate mode	(see file)
/refactor	Refactor proposal	(see file)
/tests	Tests mode	(see file)
/doc	Docs mode	(see file)

Skills (auto-discovered)

These trigger automatically based on context — no slash command needed.

Skill	Mode	Trigger
security-audit	OWASP / secrets / CVE scan	Code touches auth, user input, or external APIs
error-trace	Traceback analysis and log correlation	User pastes a traceback or error message
data-validate	Schema and data quality checks	Working with data files, schemas, or pipeline outputs
dep-audit	CVEs, licenses, stale dependencies	Reviewing dependencies or updating packages
pipeline-check	ETL pipeline data flow trace	Working on ETL/pipeline code
cross-dep	Monorepo dependency analysis	Import errors or cross-package issues

Convention

Commands

• Plain markdown, no frontmatter
• Mode declaration on line 1
• {{variable}} template syntax for user inputs
• Numbered Deliverable section
• Rules section (bulleted)
• Variables section documenting each template var
• Ends with "End with an Improvement Radar."

Skills

• YAML frontmatter with name and description (under 1024 chars)
• Description includes what it does AND when to trigger
• Body content follows same convention as commands
• Lives in skills/<skill-name>/SKILL.md

Adding a New Command

1. Create command-name.md in commands/
2. Follow the command convention above
3. The symlink makes it immediately available as /command-name

Adding a New Skill

1. Create skills/<skill-name>/SKILL.md
2. Add YAML frontmatter with name and description
3. Follow the skill convention above
4. Claude auto-discovers it based on the description's trigger conditions