The 10-command starter kit is what you need Day 1. But TurtleWolfe’s full workflow uses 65 global commands + ~90 per-project commands. The full library lives in its own backup repo:
Repo:
~/repos/Claude_Commandz
That repo is a disaster-recovery mirror of the live ~/.claude/commands/ directory and every tracked per-project .claude/commands/ folder. It’s keyed to TurtleWolfe’s exact paths — you can’t just cp -a it into your own setup and expect everything to work. Use it as a reference when you want to see how a specific command is structured.
From Claude_Commandz/README.md:
Claude_Commandz/
├── global/
│ ├── .claude/commands/ # 65 files — mirror of ~/.claude/commands/
│ └── dotfiles/ # settings.json, statusline-command.sh
└── repos/
├── ScriptHammer/.claude/commands/ (23 files)
├── TurtleWolfe/.claude/commands/ (23 files)
├── good_prompt_bad_prompt/.claude/commands/ (14 files)
├── SpokeToWork---Business-Development/.claude/commands/ (13 files)
├── PeerMentor/.claude/commands/ (8 files)
├── TranScripts/.claude/commands/ (7 files)
├── KDG/.claude/commands/ (1 file)
└── drupal-v2-sandbox/.claude/commands/ (1 file)
Totals: 65 global commands + 2 dotfiles + 90 project commands across 8 repos = 157 backed-up files.
Priming / context — prep, prime, prime_repositories
SpecKit — constitution, specify, clarify, plan, tasks, analyze, implement, plus the speckit.* variants
Git / shipping — commit, ship
Quality — code-review, security-audit, secrets-scan
Testing — test, test-a11y, test-components, test-hooks, test-fail
Orchestration (Chapter 05) — dispatch, queue, queue-check, review-queue, status, next, log, refresh-inventories
Wireframe pipeline (Chapter 05) — wireframe, wireframe-plan, wireframe-prep, wireframe-focused, wireframe-fix, wireframe-review, wireframe-inspect, wireframe-screenshots, wireframe-status, hot-reload-viewer, viewer-status
Governance / RFCs (Chapter 05) — rfc, rfc-vote, vote-now, council, broadcast, memo, audit, constitution
Session — session-stats, session-summary, clean-start
Utilities — clean-transcript, extract-linkedin, rpg_subsystem_scaffold, read-spec, read-issues, secrets-scan
You will recognize maybe 15 of these from the starter kit + Chapter 05. The rest are situational, one-off, or TurtleWolfe-specific.
If you want one of TurtleWolfe’s commands:
~/repos/Claude_Commandz/global/.claude/commands/~/.claude/commands/<name>.md (global) or <project>/.claude/commands/<name>.md (per-project)Do NOT copy commands wholesale without reading them first. Some commands (/dispatch, /queue) assume the 27-terminal tmux infrastructure. Some (/wireframe) assume a specific SVG validator script. They won’t work in your repo without the supporting cast.
The pattern from Claude_Commandz/README.md:
# In your own backup repo
mkdir -p global/.claude/commands global/dotfiles
# Refresh: copy from live system
cp -a ~/.claude/commands/. global/.claude/commands/
cp ~/.claude/settings.json global/dotfiles/settings.json
# Per-project (loop over existing repos)
for d in repos/*/; do
name=$(basename "$d")
src=~/repos/"$name"/.claude/commands
[ -d "$src" ] && cp -a "$src/." "$d/.claude/commands/"
done
# Commit
git add -A
git commit -m "refresh: sync command backups"
Run this periodically. Secrets stay out (settings.json doesn’t contain any — enforce this), per-project customizations stay in. If your machine dies, clone the backup, run the restore in reverse, and you’re back in business.
~/repos/Claude_Commandz/README.md — the full backup repo