review

claude-skills

Skills and plugins for Claude Code.

Install

As a plugin (recommended — gets you the multi-agent skills like /review and /quorum):

claude plugin add drewdrewthis/claude-skills

Individual skills (copy a single skill into your global skills directory):

cp -r <skill-name> ~/.claude/skills/

Skills

Standalone

Skill	Description
/create-skill	Turn a workflow into a well-structured Claude Code skill
/evolve-skill	Improve a skill based on how it just performed — patches friction, corrections, and missed steps
/deep-reflect	Audit Claude session history (JSONL logs) for mistakes, failed patterns, and token waste
/retro	Weekly engineering retrospective with commit analysis, quality metrics, and trend tracking
/implement	Start implementation of a GitHub issue (/implement #123)
/drive-pr	Drive a PR to mergeable state — fix CI failures, address review comments, loop until green
/write-pr	Create a PR with a filled-in template

Plugins

Plugins bundle skills with dedicated agents and reference docs. Install them via claude plugin add.

review (v2.0)

Parallel code review with 4 specialized agents running concurrently, followed by persona-driven perspectives.

Agents:

• principles-reviewer — SRP, design, readability, simplicity, extensibility
• hygiene-reviewer — reuse, patterns, idioms, dead code, bloat
• test-reviewer — test pyramid placement, coverage, naming
• security-reviewer — PII, secrets, data exposure

Bundled docs: CODING_STANDARDS.md, TESTING_PHILOSOPHY.md

/review              # review current diff
/review src/auth/    # focus on specific files

quorum (v1.0)

Debate a design question with N agents arguing different positions, then synthesize a recommendation. Includes a devil's advocate agent.

/quorum "Should we use Redis or Postgres for our job queue?"
/quorum "Monorepo vs polyrepo for our microservices"

Structure

claude-skills/
├── .claude-plugin/
│   └── marketplace.json       # plugin registry
├── create-skill/SKILL.md
├── deep-reflect/
│   ├── SKILL.md
│   └── scripts/               # session analysis helpers
├── drive-pr/SKILL.md
├── evolve-skill/SKILL.md
├── implement/SKILL.md
├── retro/SKILL.md
├── write-pr/
│   ├── SKILL.md
│   └── docs/PR_TEMPLATE.md
└── plugins/
    ├── review/
    │   ├── .claude-plugin/plugin.json
    │   ├── agents/            # 4 reviewer agents
    │   ├── docs/              # coding standards, testing philosophy
    │   └── skills/review/SKILL.md
    └── quorum/
        ├── .claude-plugin/plugin.json
        ├── agents/            # devil's advocate
        └── skills/quorum/SKILL.md

Writing your own skills

Use /create-skill to scaffold a new skill, or see any SKILL.md in this repo for the format. After using a skill, run /evolve-skill <name> to tighten it based on what actually happened.

License

MIT