| name | review |
| description | This skill should be used when the user asks to 'review changes', 'review my code', 'run a code review', 'review last commit', 'review recent changes', or 'check code against spec'. |
Code Review Orchestrator
Dispatch the engy:reviewer agent to simplify code directly, then surface severity-ordered findings.
Inputs
- Scope: user-specified → arguments → auto-detect (uncommitted → last commit → branch diff)
- Spec (optional): only if user explicitly provides a spec/plan path for alignment checking
Process
Step 0: Determine Scope
Resolve the review scope. Show a summary: files changed, lines added/removed, directories affected.
Determine what are the features being implemented. Pass file paths to any relevant plans or tasks to the review agent.
Resolution order:
- User-specified commit range, file list, or branch comparison
- Arguments passed when invoked (e.g., "last commit", "staged changes")
- Auto-detect: uncommitted changes (
git diff HEAD) → last commit (git diff HEAD~1..HEAD) → branch diff against default branch
Step 1: Dispatch engy:reviewer
Spawn the engy:reviewer agent via the Agent tool:
Agent tool:
subagent_type: engy:reviewer
mode: bypassPermissions
prompt: |
Review the following files changed in [scope description]:
Changed files:
- [list of file paths from the diff]
Diff summary:
[paste the git diff output or key changes]
Project conventions: [path to CLAUDE.md if available]
Features: [path to any plan files, tasks or features lists user provided one, otherwise omit]
Run both phases (Simplify then Review) on these files.
If plan/spec files are provided, also validate requirements coverage:
- Check each referenced FR in the plan
- Confirm the implementation actually fulfills it end-to-end, not just that code exists
- Report any FRs that are missing, only partially implemented, or not covered by tests
The agent runs two phases internally:
- Simplify — direct code changes, no behavior modifications, no user approval
- Review — surface findings tagged with severity and file:line
Step 2: Verify Build
After the agent completes, run the project build/test command (discovered from CLAUDE.md, package.json, or Makefile).
If simplification broke the build: agent fixes (2 attempts max), then reverts simplification and keeps review findings.
Step 3: Present Results
Format the agent's output into the report below. Number all findings, sorted Critical → High → Medium.
Output Format
## Code Review: [scope description]
### Simplified
[Summary of direct changes made, or "No simplifications" / "Reverted due to build failure"]
### Issues
1. **[CRITICAL]** `file:line` — Description — Suggested fix: ...
2. **[HIGH]** ...
3. **[MEDIUM]** ...
### Requirements Coverage
[Only if plan/spec was provided. For each FR: status (covered / partial / missing) and evidence.]
### Summary
[2-3 sentences: assessment, severity counts, recommendation]
Severity
- Critical — Breaks correctness, security vulns, data loss
- High — Architectural violations, missing error handling, wrong dependency direction
- Medium — Pattern inconsistencies, naming, readability, minor test gaps
Key Principles
- Orchestrate, do not review — dispatch the agent, format results
- Simplification is autonomous — no approval needed, no behavior changes
- Single agent, single pass — saves tokens
- Build verification — if simplification breaks build, fix or revert
- Every finding: file:line + concrete suggestion
Flow Position
Previous: implement | Next: complete-project
When the code review is complete and all critical/high issues are resolved, commit the work. Once the project's tasks are all in done status, proceed with /engy:complete-project to distill knowledge and archive it.
