review
Validate plans, execution, or PRs against wish criteria — returns SHIP / FIX-FIRST / BLOCKED with severity-tagged gaps.
Best use case
review is best used when you need a repeatable AI agent workflow instead of a one-off prompt.
Validate plans, execution, or PRs against wish criteria — returns SHIP / FIX-FIRST / BLOCKED with severity-tagged gaps.
Teams using review should expect a more consistent output, faster repeated execution, less prompt rewriting.
When to use this skill
- You want a reusable workflow that can be run more than once with consistent structure.
When not to use this skill
- You only need a quick one-off answer and do not need a reusable workflow.
- You cannot install or maintain the underlying files, dependencies, or repository context.
Installation
Claude Code / Cursor / Codex
Manual Installation
- Download SKILL.md from GitHub
- Place it in
.claude/skills/review/SKILL.mdinside your project - Restart your AI agent — it will auto-discover the skill
How review Compares
| Feature / Agent | review | Standard Approach |
|---|---|---|
| Platform Support | Not specified | Limited / Varies |
| Context Awareness | High | Baseline |
| Installation Complexity | Unknown | N/A |
Frequently Asked Questions
What does this skill do?
Validate plans, execution, or PRs against wish criteria — returns SHIP / FIX-FIRST / BLOCKED with severity-tagged gaps.
Where can I find the source code?
You can find the source code on GitHub using the link provided at the top of the page.
Related Guides
AI Agents for Coding
Browse AI agent skills for coding, debugging, testing, refactoring, code review, and developer workflows across Claude, Cursor, and Codex.
Cursor vs Codex for AI Workflows
Compare Cursor and Codex for AI coding workflows, repository assistance, debugging, refactoring, and reusable developer skills.
SKILL.md Source
# /review — Universal Review Gate Validate any artifact against its wish criteria. Dispatch as a subagent — never review your own work inline. ## Context Injection This skill receives its scope from the dispatch layer: - **Target** — what is being reviewed (wish draft, completed work, or PR diff) - **Wish path** — `.genie/wishes/<slug>/WISH.md` in the shared worktree - **Injected criteria** — acceptance criteria extracted from the wish If context is injected, use it directly. Do not re-parse for information already provided. ## When to Use - Before `/work` — validate a wish plan is ready for execution - After `/work` — verify implementation meets acceptance criteria - Before merge — check a PR diff against wish scope ## Flow 1. **Detect target** — determine what is being reviewed (wish draft, completed work, or PR diff). 2. **Select pipeline** — match target to Plan, Execution, or PR checklist below. 3. **Run checklist** — evaluate each criterion, collecting evidence. 4. **Run validations** — execute any validation commands; capture pass/fail output. 5. **Tag gaps** — classify every unmet criterion by severity. 6. **Return verdict** — one of SHIP, FIX-FIRST, or BLOCKED (see Verdicts). 7. **Write next steps** — exact fixes, files, and commands for each gap. ## Council Participation When a council team is active, the review can incorporate council perspectives: - Post review findings to team chat for council input via `genie chat post --team <team>` - Council members may surface risks or blind spots missed by the standard checklist - Council input is advisory — the verdict is still determined by the checklist ## Pipelines ### Plan Review (before `/work`) - [ ] Problem statement is one sentence and testable - [ ] Scope IN has concrete deliverables - [ ] Scope OUT is explicit — boundaries stated - [ ] Every task has testable acceptance criteria - [ ] Tasks are bite-sized and independently shippable - [ ] Dependencies tagged (`depends-on` / `blocks`) - [ ] Validation commands exist for each execution group ### Execution Review (after `/work`) - [ ] All acceptance criteria met with evidence - [ ] Validation commands run and passing - [ ] No scope creep — only wish-scoped changes - [ ] Work is auditable — commands and outcomes captured - [ ] Quality pass: security, maintainability, correctness - [ ] No regressions introduced ### PR Review (before merge) - [ ] Diff matches wish scope — no unrelated changes - [ ] File list matches wish's "Files to Create/Modify" - [ ] No secrets, credentials, or hardcoded tokens in diff - [ ] Tests pass (if applicable) - [ ] Commit messages reference wish slug ## Severity & Verdicts | Severity | Meaning | Blocks? | |----------|---------|---------| | CRITICAL | Security flaw, data loss, crash | Yes | | HIGH | Bug, major perf issue | Yes | | MEDIUM | Code smell, minor issue | No | | LOW | Style, naming preference | No | | Verdict | Condition | Next step | |---------|-----------|-----------| | **SHIP** | Zero CRITICAL/HIGH gaps, validations pass | See SHIP Next-Steps below | | **FIX-FIRST** | Any CRITICAL/HIGH gap or failing validation | Auto-invoke `/fix` | | **BLOCKED** | Scope or architecture issue requiring wish revision | Escalate to human | ### SHIP Next-Steps (context-dependent) | Review Context | On SHIP | |---------------|---------| | Plan review (after `/brainstorm`) | Proceed to `/wish` to create executable plan | | Plan review (after `/wish`) | Proceed to `/work` to execute the plan | | Execution review (after `/work`) | Create PR targeting `dev` | | PR review (before merge) | Merge to `dev` (agents) or approve for human merge | ### Auto-Invocation on FIX-FIRST When the verdict is FIX-FIRST: 1. Auto-invoke `/fix` with the severity-tagged gap list. 2. After `/fix` completes, re-run `/review` (max 2 fix loops). 3. If still FIX-FIRST after 2 loops, escalate as BLOCKED. ### Unclear Root Cause When a failure is found but the root cause is unclear: - Invoke `/trace` to investigate before dispatching `/fix`. - `/trace` produces a diagnosis report; `/fix` uses it to apply the correction. ## Dispatch **The reviewer must not be the engineer.** Always dispatch review as a separate subagent. ```bash # Spawn a reviewer subagent genie spawn reviewer ``` ## Rules - Never mark PASS without evidence — verify, don't assume. - Never ship with CRITICAL or HIGH gaps. - Never implement fixes during review — hand off to `/fix`. - Every FAIL includes actionable fix (file, command, what to change). - Keep output concise, severity-ordered, and executable.
Related Skills
work
Execute an approved wish plan — orchestrate subagents per task group with fix loops, validation, and review handoff.
wish
Convert an idea into a structured wish plan with scope, acceptance criteria, and execution groups for /work.
trace
Dispatch trace subagent to investigate unknown issues — reproduces, traces, and reports root cause for /fix handoff.
report
Investigate bugs comprehensively — cascade through /trace, capture browser evidence, extract observability data, and auto-create a GitHub issue with all findings.
refine
Transform a brief or prompt into a structured, production-ready prompt via prompt-optimizer. File or text mode.
learn
Diagnose and fix agent behavioral surfaces when the user corrects a mistake — connects to Claude native memory.
genie
Transform any Claude Code session into an Automagik Genie orchestrator — guide users through brainstorm, wish, team, and PR lifecycle.
fix
Dispatch fix subagent for FIX-FIRST gaps from /review, re-review, and escalate after 2 failed loops.
dream
Batch-execute SHIP-ready wishes overnight — pick wishes, orchestrate workers, review PRs, wake up to results.
docs
Dispatch docs subagent to audit, generate, and validate documentation against the codebase.
council
Brainstorm and critique with 10 specialist viewpoints. Use for architecture, plan reviews, or tradeoffs.
brainstorm
Explore ambiguous or early-stage ideas interactively — tracks wish-readiness and crystallizes into a design for /wish.