claude-skills

# claude-skills

> A curated collection of [Claude Code](https://claude.com/claude-code) skills for SEO/GEO improvement, code review, and software delivery workflows. Skills install via symlink into `~/.claude/skills/` and are reusable across any project.

Skills live as directories containing a `SKILL.md` with YAML frontmatter. The install script symlinks each directory into `~/.claude/skills/` so edits propagate instantly — no rebuild, no reload.

## Skill teams

### SEO team — end-to-end SEO + GEO improvement

- [seo-audit](https://github.com/vatsal2210/claude-skills/blob/main/seo-audit/SKILL.md): End-to-end SEO + GEO orchestrator. In deep mode, dispatches 4 parallel subagents (schema coverage, technical sweep, content/GEO, crawler simulation) via the Claude Code Agent tool and synthesizes a unified report. Supports Quick, Deep, and Team modes.
- [seo-measure](https://github.com/vatsal2210/claude-skills/blob/main/seo-measure/SKILL.md): Weekly compounding measurement loop. Pulls Google Search Console data, runs live AI citation checks (ChatGPT, Claude, Perplexity), diffs against last week, and writes findings to per-site memory. The learning system that makes the SEO team site-specific over time.

### Review team — plan → build → ship pipeline

- [plan-review](https://github.com/vatsal2210/claude-skills/blob/main/plan-review/SKILL.md): Review a plan document for gaps, risks, and missing edge cases before implementation starts.
- [gap-check](https://github.com/vatsal2210/claude-skills/blob/main/gap-check/SKILL.md): Post-implementation review covering both code quality gaps and dead code cleanup.
- [simplify](https://github.com/vatsal2210/claude-skills/blob/main/simplify/SKILL.md): Review changed code for reuse, over-engineering, and clarity — then fix approved issues.
- [pre-ship](https://github.com/vatsal2210/claude-skills/blob/main/pre-ship/SKILL.md): Production readiness checklist across seven categories (env, DB, backwards compat, errors, security, monitoring, cleanup) before deploying.
- [pr-triage](https://github.com/vatsal2210/claude-skills/blob/main/pr-triage/SKILL.md): Fetch GitHub PR review comments and triage each as bug, improvement, by-design, or needs-discussion.

## Install

```bash
git clone https://github.com/vatsal2210/claude-skills.git ~/developer/claude-skills
cd ~/developer/claude-skills
./install.sh                               # install all
./install.sh --list                        # list available skills
./install.sh seo-audit seo-measure         # install only the SEO team
```

## Design patterns worth borrowing

This repo is also a worked example of two reusable patterns:

1. **Orchestrator + parallel subagents** — one skill dispatches multiple Agent tool calls in a single message for parallelism, then synthesizes on the main thread. `seo-audit` is the reference implementation.
2. **Measurement + per-site memory** — a separate skill runs on a recurring schedule, pulls metrics, compares against history, and writes findings to a per-target memory directory so knowledge compounds over time. `seo-measure` is the reference implementation.

Both patterns can be adapted to performance audits, security reviews, accessibility audits, cost reviews, or any domain where a one-shot audit + ongoing measurement loop makes sense.

## About

- License: MIT
- Repository: https://github.com/vatsal2210/claude-skills
- Built for Claude Code, works with any project type, no external dependencies beyond `git`, `curl`, and `bash`.