Menu
Use when you need code review, verification, or quality checks on your work.
Installer avec Codex ou Claude Copiez ce prompt, collez-le dans Codex, Claude ou un autre assistant, puis laissez-le vérifier la page du skill et l'installer pour vous.
| name | team-shinchan:review |
| description | Use when you need code review, verification, or quality checks on your work. |
| user-invocable | true |
If args is empty or only whitespace:
Ask user: "What would you like me to review? (code changes, file paths, or describe what to check)"
STOP and wait for user response
If args length > 2000 characters:
Truncate to 2000 characters
Warn user: "Request was truncated to 2000 characters"
Quality-first: a single agent juggling all five dimensions in one context under-attends the later ones and never re-checks its own findings. When /team-shinchan:review runs in the main loop, the adversarial multi-dimension Workflow is the default — independent per-dimension agents, a non-skippable per-finding refutation, completeness + interaction critics, and a deterministic 3-lens judge panel.
scope + files (git diff), pick the rubric from ${CLAUDE_PLUGIN_ROOT}/agents/_shared/eval-rubrics.json, and resolve Action Kamen's persona via node ${CLAUDE_PLUGIN_ROOT}/src/workflow-personas.js actionkamen — exactly as team-shinchan:fierce-review Step 1 specifies. Also resolve node ${CLAUDE_PLUGIN_ROOT}/src/workflow-personas.js --learnings actionkamen and inject it as args.learnings (mirror fierce-review; distinct from args.persona — FR-5/AC-8).deepen defaults to true):
Workflow({
scriptPath: "${CLAUDE_PLUGIN_ROOT}/skills/fierce-review/fierce-review.workflow.js",
args: { scope: "<one-line scope>", files: ["<path>", "..."], baseRef: "main",
rubric: { /* object from eval-rubrics.json */ },
persona: "<string from workflow-personas.js actionkamen>",
learnings: "<string from workflow-personas.js --learnings actionkamen>" }
})
team-shinchan:fierce-review Steps 3–4 (write .shinchan-docs/reviews/REVIEW-{NNN}.json; an APPROVED artifact counts as code-review evidence for team-shinchan:verification-before-completion). Never finalize without user confirmation.workflow() throws inside a Task child, and a Workflow can't be fired from a hook (HARD INVARIANT). So only when this skill is reached by delegation into a subagent (e.g. Shinnosuke/Action Kamen invoked it) or from a hook, skip the Workflow and run the single Action-Kamen Task below — the one remaining cheap single-pass path.
Task(
subagent_type="team-shinchan:actionkamen",
model="opus",
prompt=`/team-shinchan:review has been invoked.
## Code Review Request
Perform thorough review covering:
| Category | Check Items |
|----------|-------------|
| Correctness | Logic errors, edge cases, expected behavior |
| Security | Vulnerabilities, input validation, auth issues |
| Performance | N+1 queries, memory leaks, bottlenecks |
| Code Quality | Readability, maintainability, conventions |
| Tests | Coverage, test quality, missing tests |
## Review Output Requirements
- Show review process in real-time
- List all issues found with severity (CRITICAL/HIGH -> MUST fix; MEDIUM -> SHOULD fix; LOW -> COULD fix)
- Provide specific file:line references
- Give actionable fix recommendations
- Final verdict: APPROVED ✅ or REJECTED ❌
- If REJECTED: list specific issues that must be fixed, ordered by severity
User request: ${args || '(Please describe what to review)'}
`
)
STOP HERE.
Use when you have a large-scale, multi-phase project requiring orchestrated execution.
Use when you need persistent looping until a task is fully complete.
Deterministic adversarial code review for high-stakes scope — independent per-dimension review, a non-skippable per-finding refutation, completeness + interaction critics, and a deterministic 3-lens rubric judge panel. Opt-in main-loop Workflow tier.
Use when the user wants to review accumulated skill feedback, verdict trends, or improvement candidates collected during Stage 4 retrospectives. Trigger on "show skill feedback", "스킬 피드백 보여줘", or finding which skills need /writing-skills work.
Deterministic competitive code tournament — N builders independently solve one task and return patches, an Action-Kamen judge scores them head-to-head, the winner is picked by score and applied. Opt-in main-loop Workflow tier.
Deterministic adversarial debate for high-stakes or irreversible decisions — mandatory refutation plus a scored judge panel. Opt-in main-loop Workflow tier.