Loading...
Loading...
Operator-invoked review of the full branch (every commit vs. the base branch) plus any uncommitted local changes before opening or updating a PR. Strict gatekeeper stance with per-diff and branch-level checks. Language-agnostic.
npx skill4agent add alimtunc/turkit pre-pr-reviewreferences/review-rubric.mdpre-commit-reviewpre-pr-review<base>..HEAD.turkit.yaml → base_branchgit symbolic-ref refs/remotes/origin/HEADmaingit status --shortLOCAL_DIRTY = truepre-commit-reviewpre-pr-reviewgit rev-list --count <base>..HEADgit log --oneline <base>..HEADgit log --stat <base>..HEADgit diff <base>..HEADgit diff <base>..HEAD --name-onlygit show <hash>LOCAL_DIRTYgit diff --cachedgit diffgit ls-files --others --exclude-standard<base>..HEAD.turkit.yamlrules.docsCLAUDE.mdAGENTS.mddocs/conventions/*.md.turkit.yaml → commands.lintreferences/build-tool-detection.mdLOCAL_DIRTYBranchLocalreferences/branch-review.mdHEADReady for PRreferences/branch-review.mdThe Branch summary, Per-Commit, Cross-Commit, Branch-Level, and Verdict sections live in. The per-diff sections below are specific to this skill.references/branch-review.md
## Mechanical Pre-pass (lint)
- Ran: `<exact command>`
- Findings kept: N
- Findings dropped as false positives: N — list with reasons
- Notable rules triggered: short list with file:line
## Local (uncommitted)
> Only present when the working tree was dirty. Findings against the per-diff rubric on staged + unstaged + untracked.
- [P0|P1] [Category] [Local] [file:line] What and why
- Or: "No findings — local diff is clean and ready to commit."
## Fixes Applied
> Auto-fixes from the shared rubric. All landed **unstaged** on current HEAD.
- [Category] [file:line] What changed
## Required Changes
- [P0|P1] [Category] [file:line | commit:<hash>] What must change and why it cannot be auto-fixed
- Suggested rebase plan, if history rewrite is required (do not execute it)
## Blocking Issues
- [P0|P1] [Category] [file:line | commit:<hash>] Rare issue not covered above
## Suggested Improvements
- [Category] Improvement and expected benefit
## Positive Signals
- Short bullet only when it reduces review ambiguity
## Verification
- Ran: exact commands
- Residual failures, if any
- Skipped: relevant checks not run and why
## Residual Risks
- Remaining uncertaintyreferences/output-preferences.md