ワンクリックで
team-shinchan-review
Use when you need code review, verification, or quality checks on your work.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
メニュー
Use when you need code review, verification, or quality checks on your work.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
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.
| 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.