CherryHQ/cherry-studio / .agents/skills/gh-pr-review
gh-pr-review
UnverifiedAutomated Cherry Studio review for local branches, PRs, commits, files, architecture docs, and repository skills. Use for code or documentation reviews that need project-specific naming, main/renderer/shared placement and dependency rules, IpcApi and DataApi boundaries, lifecycle/service ownership, renderer hooks, React/UI conventions, and tests. Review depth adapts to diff size and runtime subagent capability (single-agent or multi-agent reviewer-verifier). Report-only by default; code fixes and GitHub submission each require explicit invocation-time authorization (`fix` / `submit`). Normal-review prompts and safe interruption behavior follow the interaction contract below. To diagnose gaps in the skill after a review session, run `/gh-pr-review diag`.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 15.4 KB |
| agents/openai.yaml | config | 247 B |
| references/checklist-evolution.md | markdown | 4.7 KB |
| references/cherry-review-guidance.md | markdown | 40.0 KB |
| references/code-checklist.md | markdown | 10.3 KB |
| references/consumer-review.md | markdown | 7.4 KB |
| references/diagnosis.md | markdown | 2.6 KB |
| references/doc-checklist.md | markdown | 3.8 KB |
| references/judgment-matrix.md | markdown | 6.8 KB |
| references/local-review.md | markdown | 6.2 KB |
| references/pr-review.md | markdown | 19.7 KB |
| references/teams-review.md | markdown | 19.0 KB |
references/local-review.md
View on GitHub# Local Review
Single-agent review for a small scope, or as the explicit fallback when a
runtime cannot launch independent subagents. Scope and `SMALL_SCOPE` are
derived by `SKILL.md` § Scope derivation. Reviews the diff and reports
confirmed issues with fix guidance; edits code only when the invocation
explicitly authorized fixing.
Follow `SKILL.md` § Interaction and interruption contract; this flow introduces
no additional prompt category.
## Input from SKILL.md
- `REVIEW_TARGET`, the resolved review scope, and `SMALL_SCOPE`, all derived
by `SKILL.md` § Scope derivation.
- `AUTHORIZED_FIX`: `true` only when the invocation explicitly granted
fixing (`fix` modifier or equivalent user wording). Commit and range
targets are always report-only regardless of the flag.
- `LIMITED_SINGLE_AGENT`: `true` only when a non-small scope was routed here
because the runtime has no subagent capability. Default `false`.
## References
| File | Purpose |
|------|---------|
| `consumer-review.md` | Consumer review stage (changes adding/expanding shared surface) |
| `code-checklist.md` | Code review checklist |
| `doc-checklist.md` | Document review checklist |
| `cherry-review-guidance.md` | Cherry Studio project-specific review boundaries |
| `judgment-matrix.md` | Risk levels, worth-fixing criteria, special rules |
| `checklist-evolution.md` | Checklist update flow and rules |
---
## Step 1: Scope
Scope, resolved-scope emptiness, and `SMALL_SCOPE` are owned by `SKILL.md`
§ Scope derivation. The router already resolved them; when invoked
standalone, derive them with those rules and print its usage examples if the
resolved scope is empty. Do not restate the derivation here.
---
## Step 2: Review
Run the review stages defined in `SKILL.md` § Review Stages in order.
1. **Product Demand gate** — follow `SKILL.md` § Review Stages, stage 1 for
the semantic-impact, decision-evidence, and mode rules. Preserve its three
states: no product impact, established direction, and open product decision.
Only an open product decision may prompt in an interactive run or use
record-only behavior in an explicitly automated run. Skip this step when
the PR wrapper already ran the gate.
2. **Consumer review** — whenever the diff adds or expands shared surface,
judged by diff semantics rather than change label. Follow
`consumer-review.md`; only surviving surfaces continue to the stages below.
3. **Architecture-First, Implementation, Style** — as follows.
Review the diff. Apply `code-checklist.md` to code files,
`doc-checklist.md` to documentation files. Apply `cherry-review-guidance.md` to
code, mixed, Cherry architecture documentation, and project-skill changes:
first read the docs its "Mandatory Baseline Docs" section requires for the
touched processes, then load only the on-demand references it routes to.
Review architecture-first — settle placement, ownership, and
abstraction-integrity findings against those docs before line-level detail;
doc violations are Warning minimum. For React component changes, also consult
`vercel-react-best-practices` skill for detailed performance patterns. When
changed lines depend on surrounding context, read the relevant sections or
related definitions as needed. Untracked files have no diff — review their
full contents as new code.
If the branch has an associated GitHub PR, inspect its checks with `gh pr
checks` and include failing or pending CI in the review. Do not run `pnpm lint`,
`pnpm test`, or `pnpm format` locally during review. If no associated PR exists,
state that CI validation is unavailable and keep the result explicitly limited
to static review.
For each issue found:
- Provide a code citation (file:line + snippet) from the current tree.
- Self-verify by re-reading the code — confirm or withdraw.
- If a cited path/line no longer exists, locate the correct file/path via `git diff --name-only` or file search before reporting.
**Output rule**: only present the final confirmed issues to the user. Do not
output analysis process, exclusion reasoning, or issues that were considered
but ruled out.
---
## Step 3: Filter
Consult `judgment-matrix.md` for risk level assessment, worth-fixing criteria,
and special rules. Discard issues that are not worth reporting.
If no issues remain after filtering, keep an empty issue list and continue to
Step 4 so every mandatory disclosure is still reported. Step 5 may be skipped
because there are no confirmed issues to evolve into checklist candidates.
---
## Step 4: Fix and report
Do not ask the user which issues to fix.
Apply `judgment-matrix.md` § Handling by Risk Level, which owns what each risk
level permits under `AUTHORIZED_FIX`. Keep every applied fix at the defect's
altitude per `cherry-review-guidance.md` § Fix Recommendation Policy.
Present a summary of what was reviewed and either the issues fixed/reported
with their fix guidance or "no issues found". If
`LIMITED_SINGLE_AGENT = true`, explicitly state that the scope was non-small
but the runtime had no subagent capability, so the review was single-agent and
did not include independent adversarial verification. In an automated session
with an open product decision, include the Product Demand summary — impact,
direction, and points needing human confirmation — explicitly marked as
awaiting a product decision, never as approved.
Validation: when fixes were applied, the session is a coding task — run the
validation selected per `SKILL.md` § Validation after applied fixes and report
the results; a failure caused by a fix means the fix is reverted or reported
as failed.
When nothing was edited, do not run local lint/test/format and state the CI
baseline recorded in Step 2, matching it: with an associated PR, existing CI
covers the reviewed commit but not any unpushed change; with no associated PR
— file, commit, or local-branch review — CI validation is unavailable and the
result is static review only. Never claim CI coverage the target does not have.
---
## Step 5: Checklist evolution
Review all confirmed issues from this session. If any represent a recurring
pattern not covered by the current checklist, read `checklist-evolution.md` and
record valid candidates as `proposed` in the report. A regular review never
accepts, inserts, or claims to persist checklist rules.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| finishing-a-development-branchobra/superpowersUse when implementation is complete, all tests pass, and you need to decide how to integrate the work | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| systematic-debuggingobra/superpowersUse when encountering any bug, test failure, or unexpected behavior, before proposing fixes | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| test-driven-developmentobra/superpowersUse when implementing any feature or bugfix, before writing implementation code | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| skill-creatoranthropics/skillsCreate new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, edit, or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |