Claude Global Skills

claude-global-skills

A curated collection of [Claude Code](https://claude.com/claude-code) **skills** — a multi-agent code-review loop, exhaustive local chat-history search, an evidence-grounded session end/resume pair, Gemini image generation, and rigorous transcript retrospectives — that run **machine-wide with nothing but `python` and `node`**. No per-project install: drop a skill into `~/.claude/skills/` and invoke it as a slash command in any session.

These are global skills, version-controlled here so they can be reviewed, shared, and deployed to a fresh machine. **The repo is the source of truth**; each live copy is a deployed (materialized) copy. Most skills target **Claude Code** (`~/.claude/skills/`); a skill can also declare **Codex** (`~/.codex/skills/`) as a target and deploy to both from one shared source — see [Cross-runtime targets](#cross-runtime-targets-claude-code--codex).

Quickstart

Install a single skill by copying its directory into your Claude Code skills folder:

git clone https://github.com/BryceEWatson/claude-global-skills
cp -r claude-global-skills/review-loop ~/.claude/skills/

Then invoke it in any Claude Code session:

/review-loop

That's it for prompt-only skills. A few skills also ship an installer (they wire a hook into `~/.claude/settings.json`) — run `node ~/.claude/skills//install.cjs` after copying. See each skill's section below.

Skills

Core — portable, useful for anyone

Skill	What it does
`review-loop`	Dispatches a multi-agent review team over your session's diff, runs an execution-grounded lint/test/build check, validates each finding through a falsifier stage, and posts a commit-pinned verdict on the PR. Four lens sets by artifact type: code, plans, analytical claims, and finished deliverables (whether the human it's handed to can actually use it). Ships a Stop-hook + installer.
`gemini-image`	Generate and edit images via