review

📁 kelvinz/cobb 📅 6 days ago

总安装量

周安装量

#16043

全站排名

安装命令

npx skills add https://github.com/kelvinz/cobb --skill review

Agent 安装分布

amp 8

opencode 8

kimi-cli 8

codex 8

claude-code 8

Skill 文档

review

Review one change set in one mode and return a decision-led report.

Guardrails

Require an explicit mode: pr or local.
Do not implement or modify code.
Do not commit, merge, push, or delete branches.
Do not update PRD/todo tracking files here.
Update tasks/memory.md only for durable review outcomes (recurring risks, release-critical gotchas, or confirmed follow-up decisions).
Do not invent test results; run checks or call out missing evidence.

Inputs

review mode:
- pr: PR URL/number (preferred)
- local: base branch (default: repository default branch resolved from origin/HEAD; ask if unclear)
optional PRD path (if scope validation is needed)

Workflow

Confirm mode and target.
Collect context:
- pr mode:
  - Preferred (gh available):
    - gh pr view --json url,number,title,body,state,isDraft,baseRefName,headRefName,files,additions,deletions
    - gh pr diff
    - gh pr checks
  - Fallback (gh unavailable/auth fails):
    - ask for base/head branch names if not already provided
    - git fetch origin "<base>" "<head>"
    - git diff "origin/<base>...origin/<head>"
    - git log "origin/<base>..origin/<head>" --oneline
    - mark CI/check status as Missing evidence unless the user provides CI artifacts
- local mode:
  - git diff "<base>...HEAD"
  - git log "<base>..HEAD" --oneline
  - git status --short
Compare the change set against required behaviour:
- correctness and edge cases
- security risks and data handling
- test depth and regression risk
- scope control (especially if PRD path is provided)
  - Compare diff vs PRD ‘In scope’ and completed user stories; flag any diff not attributable to a PRD requirement.
Classify findings:
- blockers (must fix)
- suggestions (optional improvements)
- missing evidence (tests/checks not run, unclear behaviour)
  - If unable to run checks (CI-only, permissions), mark as ‘Missing evidence’ and request the specific artifact (CI link, log, or command for user to run).
Produce the report with a clear recommendation:
- pr mode: Ready to accept PR: Yes or Ready to accept PR: No
- local mode: Good to commit: Yes or Good to commit: No
- if decision is No, include explicit fix items and ask the user to address them before rerunning review in the same mode
Evaluate memory-worthy review outcomes and update tasks/memory.md inline when needed:
- systemic risks likely to recur
- key security or data-handling decisions
- durable follow-up decisions that affect future work
- if no durable outcome exists, mark memory as skipped with reason in the report

Review Checklist

Correctness:
- empty/null/error paths
- boundary values and state transitions
- ordering/concurrency/time assumptions (if applicable)
Security:
- authn/authz behaviour
- input validation and output encoding
- secret/PII handling and logging safety
- dependency risk for newly introduced packages
Tests and verification:
- happy path + key failure paths
- regression coverage in touched areas
- manual verification steps when automation is missing
Maintainability:
- naming clarity and control-flow simplicity
- comments/docs for non-obvious decisions only

Report Template

PR Mode Report
Mode: pr
Decision:
- Ready to accept PR: Yes | No

Blockers (must fix):
- â¦

Suggestions (nice to have):
- â¦

Missing evidence:
- â¦

Security notes:
- â¦

Regression risks / watch-outs:
- â¦

Memory updates:
- Updated: <summary> | Skipped: <reason>

Recommended next step:
- If Ready to accept PR=No: ask user to fix blockers, then rerun `review` in `pr` mode.
- If Ready to accept PR=Yes: run `commit` in `finalise` mode.

Local Mode Report
Mode: local
Decision:
- Good to commit: Yes | No

Blockers (must fix):
- â¦

Suggestions (nice to have):
- â¦

Missing evidence:
- â¦

Security notes:
- â¦

Regression risks / watch-outs:
- â¦

Memory updates:
- Updated: <summary> | Skipped: <reason>

Recommended next step:
- If Good to commit=No: ask user to fix blockers, then rerun `review` in `local` mode.
- If Good to commit=Yes: run `commit` in `commit` mode.

Output

Return the review report with explicit memory update status.
Keep the decision explicit and unambiguous.
End with a short status block:
- Files changed: list of created/updated files
- Key decisions: any assumptions or choices made (if any)
- Next step: recommended next skill or action

GitHub 仓库 ↗ ← 返回陌讯 Skills 聚合平台