Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →qf-studio/branch-audit-safe-stale-branch-investigation-and-pruning
name: branch-audit
Passed 16 of 21 test cases with the skill, and 9 of 21 without it.
and 7 more in the full report
8 fixed9 passed either way5 still failing0 broke
Tokens, turns and time describe resource use, not a dollar price.
One of 7 test cases the skill fixed. Same model, same prompt, with and without the skill.
The user marked `feat/openrouter-engine` as a sensitive branch earlier. My audit shows `feat/openrouter-engine` has a merged PR and 0 unique commits. Should I propose its deletion?
| Checked | Without skill | With skill |
|---|---|---|
| Leaves explicitly flagged sensitive branches untouched regardless of audit evidence. | Missed | Passed |
The instructions your model reads when the skill loads.
This repo accumulates local branches fast: interactive sessions, the Pilot
daemon (pilot/GH-*), recovery work, backups. A 2026-07-05 audit found 29
local branches; 21 were safely deleted — but only after content-level
verification caught, among others, 15 unpushed commits on feat/aws-bench
that existed nowhere else. Ancestry checks alone would have missed that.
Prime directive: a branch is never deleted without explicit per-branch user approval, and never on ancestry evidence alone.
git branch --format='%(refname:short)' # never pipe through head — a truncated
# list caused a missed-branch round
git worktree list # branches checked out in worktrees
# cannot be deleted and may be live sessions
Exclude from the audit: the current main checkout, any claude/* branch
attached to a worktree (possibly a live session), and pilot-worktree-GH-*
daemon worktrees (per CLAUDE.md, leave alone).
For each branch collect:
git log -1 --format='%h %cd %s' --date=short "$b" # tip
gh pr list --head "$b" --state all --json number,state # PR record
test cases passed, vs 9 / 21 without
More skills from the same source and nearby in the registry, with their headline benchmark lift.