digitaltracer/agent-skills / skills/triage-reviews
triage-reviews
VerifiedFetch all unresolved PR review comments (Cursor bot, entelligence, codex-connector, humans) via gh, verify each claim against the actual code, classify valid / false-positive / stale with proof, and apply minimal targeted fixes on approval. Use when the user wants all PR comments addressed or asks to sweep a PR's reviews. For a single pasted comment, use triage-comment instead.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 2.9 KB |
SKILL.md
View on GitHub---
name: triage-reviews
description: >-
Fetch all unresolved PR review comments (Cursor bot, entelligence, codex-connector,
humans) via gh, verify each claim against the actual code, classify valid /
false-positive / stale with proof, and apply minimal targeted fixes on approval.
Use when the user wants all PR comments addressed or asks to sweep a PR's reviews.
For a single pasted comment, use triage-comment instead.
disable-model-invocation: true
---
# Triage PR Reviews
Replaces the paste-one-comment-at-a-time ritual. Works on the current branch's PR by default; accepts a PR number or URL as an argument. For a single pasted finding, use `/triage-comment`.
## Step 1 — Fetch every unresolved comment
```bash
gh pr view --json number,url,title 2>/dev/null # resolve PR from current branch
gh api repos/{owner}/{repo}/pulls/{n}/comments # inline review comments (bots post here)
gh api repos/{owner}/{repo}/issues/{n}/comments # issue-level comments
gh pr view {n} --json reviews # review bodies
```
Collect everything not yet resolved/outdated. Note the author (cursor bot, entelligence-ai, chatgpt-codex-connector, human) and timestamp.
## Step 2 — Verify before believing
For each finding, trace the claim through the actual code. **Bots are frequently wrong or stale — verification is the point of this skill, not ceremony.**
- **Valid** — the defect is real; cite the code path that proves it.
- **False positive** — the claim doesn't hold; show why (the invariant that protects it, the version that ships the fix, the config that makes it safe).
- **Stale** — already fixed by a later commit on the branch; name the commit.
- **Nit** — real but cosmetic; fix only if trivial.
Never fix anything you haven't verified.
## Step 3 — Verdict table
| # | Source | Location | Claim | Verdict | Proof | Proposed fix |
Order: valid first. Keep claims and proofs to one line each; the user decides from this table.
## Step 4 — Fix
On approval (or immediately when invoked with `--fix`):
- **Minimal, targeted fixes only** — exactly what the finding requires. If the review asked for one line, change one line; do not "improve" neighbors.
- Before each fix, check it introduces no regression or behavior change beyond the finding.
- Commit atomically by concern, per global commit rules. Push only when asked — except under the babysit loop below, where push is part of the cycle.
- Where useful, draft reply comments (fix reference for valid ones, a short justification for false positives).
## Babysit loop
While a PR is open and bots keep posting, run this skill on a loop instead of manually re-checking:
```
/loop 10m /triage-reviews --fix
```
Each round: fetch comments newer than the last round, triage, fix, commit, push, then check `gh pr checks`. If checks are failing, diagnose per `/fix-ci`. Stop the loop when checks are green and no unresolved comments remain.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |