test-toollisted

# Test Tool Evaluates a third-party tool from the Claude Code ecosystem with a reproducible, low-risk protocol. Captures the verdict where it counts (memory + watchlist + Tools tested catalog on the site + case study if the tool taught us something significant). ## When to use - ✅ User says "on teste X", "let's try X", "/test-tool", or shares a GitHub repo / npm package URL with intent to evaluate. - ✅ A tool already in `claude-code-codex/frontend/src/data/tools-tested.ts` with status `to-test` or `browsed` needs a real verdict. - ✅ Periodic re-test of a watchlisted tool (cf `watchlist_tools_to_revisit.md`) that hit a maturity milestone. ## When NOT to use - ❌ The user is asking *about* a tool but hasn't decided to test it yet — answer the question without spinning up the protocol. - ❌ The tool requires deep integration (auth, API keys, paid tier, multi-day setup) — the 8-step protocol assumes a quick evaluation, not a full migration. - ❌ The tool doesn't target Claude Code / the Anthropic ecosystem — out of scope. ## The 8-step protocol Track each step with TodoWrite from the start. ### Step 1 — Snapshot global state Capture md5 + listings BEFORE touching anything. This is the "control" we'll diff against post-install : ```bash echo "=== md5 baseline ===" md5sum ~/.claude/CLAUDE.md ~/.claude/settings.json echo "=== plugins/ baseline ===" ls ~/.claude/plugins/marketplaces/ ~/.claude/plugins/cache/ cat ~/.claude/plugins/installed_plugins.json | head -3 cat ~/.claude/