review-queue
/review-queue
Pokazuje, które PR-y nadal czekają na przegląd, i zestawia tabelę pewności scalenia na podstawie zapisanych dokumentów przeglądu.
- Autor
- sejas ↗
- Repozytorium
- ai-code-skills ↗
- Licencja
- MIT
- Ostatnia zmiana repozytorium
- 08.10.2026
Ocena i bezpieczeństwo
- Pliki w paczce
- 1
- Skrypty
- 0
- Komentarze HTML
- 0
- Adresy URL w SKILL.md
- 0
Brak sygnałów. Repozytorium zmienione 08.10.2026, bez odwołań do wycofanych modeli i API.
MIT, otwarta: można używać i zmieniać, z zachowaniem informacji o autorze.
Co robi i kiedy
Pokazuje, które PR-y nadal czekają na przegląd, i zestawia tabelę pewności scalenia na podstawie zapisanych dokumentów przeglądu.
Wyzwalacze z opisu
Oryginalny opis (EN)
Use when Antonio asks which PRs are still open and waiting for his review, wants a merge-confidence table from his saved review docs, or says "review queue", "rate my open PRs", "which PRs can I merge", "what's open in the repo", or invokes /review-queue.
Jak działa
- Inputs
- Steps
- Repo board (all open non-draft PRs)
- Common Mistakes
Podgląd SKILL.md
---
name: review-queue
description: Use when Antonio asks which PRs are still open and waiting for his review, wants a merge-confidence table from his saved review docs, or says "review queue", "rate my open PRs", "which PRs can I merge", "what's open in the repo", or invokes /review-queue.
---
# Review Queue
Two outputs, always both:
1. **Rated queue** — open PRs *awaiting the user's review*, rated 0-100 on merge-readiness from the review docs saved in the Obsidian reviews folder.
2. **Repo board** — *every* open non-draft PR in the repo, with age, requested reviewers and conversation state, so the user can pick something to review even when it was never assigned to them.
## Inputs
- Reviews folder: `~/claude.nosync/reviews/` — where `/quick-pr-review` writes.
- `REVIEW_REPO` — the `owner/repo` to query (your team's main repo). Review filenames may reference other repos; only rate what `gh` can reach.
## Steps
1. **Get open PRs awaiting review** (authoritative filter — do NOT infer from review docs alone):
```bash
gh pr list --repo "${REVIEW_REPO}" --search "review-requested:@me" --state open --limit 100 \
--json number,title,isDraft,reviewDecision,author,updatedAt
```
2. **List review docs**: `ls` the reviews folder. Filenames follow `YYYY-MM-DD-<repo>-<PR>-<slug>.md` (e.g. `2026-07-22-studio-4301-...md`). Prefer `.md` over `.html`.
3. **Match**: intersect open PR numbers with review-doc PR numbers. Multiple docs per PR → read only the **latest by date**.
4. **Read matched docs** (in parallel). Extract: verdict (Approve / Comment / Request changes), critical issues, suggestions with their stated confidence, positive notes, open discussion threads.
5. **Rate 0-100** merge-confidence per PR:
- Base by verdict: Approve 85-95, Comment 50-70, Request changes 10-40.
- Subtract for: high-confidence bug findings that block merge (-10 to -20), unresolved product/discussion questions (-5 to -10), large surface (20+ files, -5), draft status (-10), staleness >1 month since last update (-10, likely conflicts).
- Add for: tiny scope, screenshots + thorough tests, only optional/non-blocking findings.
6. **Output table**, sorted rating descending. Columns — strengths and weaknesses MUST be separate columns:
| PR | Title | Rating | Strengths | Weaknesses |
|---|---|---:|---|---|
- PR column: markdown link to the GitHub PR.
- Strengths/Weaknesses: 1-2 short phrases each, from the review doc; note verdict + draft/stale status in Weaknesses when applicable.
7. **List leftovers**: open PRs awaiting review with **no** review doc — name them below the table (numbers + short titles), state they can't be rated, suggest `/pr-review` or `/quick-pr-review`.
8. Note in one line that ratings blend verdict, finding severity, and staleness.
## Repo board (all open non-draft PRs)
After the rated queue, always output a second table covering **every** open non-draft PR in the repo — not just the ones assigned to the user. Two GraphQL pages of 50 cover it; merge them.
```bash
gh api graphql -f query='
query($owner:String!,$name:String!,$cursor:String){
repository(owner:$owner,name:$name){
pullRequests(states:OPEN,first:50,after:$cursor,orderBy:{field:UPDATED_AT,direction:DESC}){
pageInfo{hasNextPage endCursor}
nodes{
number title isDraft url createdAt updatedAt author{login} reviewDecision
additions deletions changedFiles
labels(first:10){nodes{name}}
reviewRequests(first:10){nodes{requestedReviewer{... on User{login} ... on Team{name}}}}
comments(last:1){totalCount}
reviews(first:50){totalCount nodes{author{login} state}}
reviewThreads(first:100){nodes{isResolved isOutdated}}
}
}
}
}' -F owner="${REVIEW_REPO%/*}" -F name="${REVIEW_REPO#*/}"
```
Filter `isDraft==false`, sort by `updatedAt` descending, and render:
| PR | Title | Author | Created | Updated | Review requested | Conversation | Status |
|---|---|---|---|---|---|---|---|
- **Created / Updated**: absolute date + age in days (`2026-07-24 · 3d`). Flag `updatedAt` older than 30 days as stale.
- **Review requested**: logins/teams from `reviewRequests` (a team name usually means an on-call rotation, not a person); `—` when nobody is on the hook — those are the orphans worth grabbing.
- **Conversation**: `unresolved` review threads first (that is the real "ongoing discussion" signal), then issue-comment count, then submitted review states (`CHANGES_REQUESTED`, `APPROVED`, `COMMENTED`). `—` when silent. Bot-only chatter (dependabot, CI) is noise — say so rather than counting it as discussion.
- **Status**: `reviewDecision` plus size (`+adds/-dels, N files`) and notable labels (e.g. `dependencies`, `breaking-change`, `Proof of Concept`).
Group or mark separately:
- PRs **authored by the user** (their own `gh` login) — their own, not review targets.
- **Dependabot** PRs — batch them into one line, don't list each.
- PRs with `APPROVED` and no unresolved threads — merge candidates someone should land.
## Common Mistakes
- Rating PRs from review docs whose PR already merged/closed — always filter through `gh` first.
- Reading every review doc — only read docs matching open+awaiting PRs (folder has 150+ files, most merged).
- Merging strengths and weaknesses into one "Reason" column — they are separate columns.
- Reading an older duplicate doc when a newer one exists for the same PR.
- Treating `review-requested:@me` results as needing re-check — that search already excludes PRs the user has reviewed.
- Skipping the repo board because the rated queue "looks complete" — the board is the point when the user wants something to pick up.
- Listing draft PRs on the board. Drafts are excluded; mention only the count.
- Counting dependabot/CI comments as an ongoing conversation.
Wersje
| Repozytorium | Wynik | Licencja |
|---|---|---|
| Ta strona · sejas/ai-code-skills | 100 | MIT |
| zahidakhyar/skills | 100 | MIT |
| Esposter/Esposter | 100 | Apache-2.0 |
Migawka bazy z 13.11.2026 · wersja paczki cfe3091e8dce (skrót SHA-256 całej paczki)
Pobrania w ostatnich 30 dniach: 0 · ranking
Zgłoś problem z tym skillem: szkodliwa treść, błędny opis albo prośba autora o usunięcie.
Podobne skille
- 100review-sweep
Przegląda wszystkie PR-y czekające na recenzję, uruchamia dla nich agentów przeglądu i sprawdza ponownie zaktualizowane PR-y. - 100merge-open-prs
Przegląda, weryfikuje (domyślnie w Dockerze) i automatycznie scala otwarte PR, gdy przechodzą CI, recenzje i lokalna weryfikacja, do 10 PR w jednym uruchomieniu. - 85review-prs
Wstępnie przegląda otwarte pull requesty z diffu i decyduje o scaleniu lub zamknięciu na GitHubie. - 96merge-open-prs
Scala wszystkie otwarte pull requesty albo wskazany zestaw PR-ów razem, w jednym przebiegu. - 84resolve-all-prs
Rozwiązuje problemy na wszystkich otwartych PR użytkownika równolegle: komentarze z przeglądu i błędy CI, tak by każdy był gotowy do scalenia. - 100ha-merge-queue
Znajduje otwarte pull requesty Home Assistant naprawdę gotowe do scalenia, sprawdzając CI, statusy bramek, akceptację code ownera, konflikty, żądane zmiany i nierozwiązane wątki.