tracking-test-caseslisted
Install: claude install-skill punkaze/skills
# tracking-test-cases
A pre-PR manual QA checklist, scoped to the current git branch, persisted globally (survives worktree deletion after merge), diffing each run against the last one so you can see what changed after a fix.
## Run loop
1. **Resolve identity.** Run `python3 "${CLAUDE_PLUGIN_ROOT}/skills/tracking-test-cases/scripts/base_branch.py"` from inside the target repo. It prints JSON: `repo`, `repoSlug`, `branch`, `branchSlug`, `baseBranch`, `baseBranchSource`, `storageDir`. If it errors (`detached HEAD` or `not inside a git repository`), stop and tell the user why — this skill needs a real checked-out branch.
2. **Load or create the case list** at `<storageDir>/cases.json`.
- If it already exists: don't overwrite it. Ask whether to just re-run the existing cases, or add more first.
- If it doesn't exist: ask the user — manual authoring, or auto-derive a draft?
- **Auto-derive:** run `git diff <baseBranch>...HEAD` (using the `baseBranch` from step 1 — never guess a different one). If a plan doc matching this branch exists under `docs/superpowers/plans/`, read it too. Draft candidate cases from what you find. **Always show the draft to the user to edit/add/remove before writing `cases.json` — never write auto-derived cases straight through unreviewed.** If the diff is empty, say so plainly and fall back to manual entry for this invocation.
- **Manual:** ask the user for cases directly.
- Write `<storageDir>/cases.json` as `{"branch", "repo", "crea