メニュー
Evaluate agent performance after a sprint, review, incident, build cycle, or multi-agent run. Use when the user asks what worked, what failed, which agents were effective, where delegation broke down, or how to improve future agent workflows.
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。
SOC 職業分類に基づく
Interactive front-end design companion — iterate on layouts, live component previews, and design tokens in a real browser, then hand the validated design to the frontend-family implement skills. Use when the user wants to design or refine UI look-and-feel, compare layout/wireframe options, preview real components, or tune color/spacing/typography before building.
Ingest, decompose, govern, and route large repository PDR packages into agent-routable work queues with manifests, decision registers, risk registers, governance gates, and swarm-ready handoff packets. Use when the user provides repo PDR packages, audit bundles, project design records, implementation plans, backlog packs, or governance artifacts and wants them parsed, scored, split into agent assignments, and dispatched through the Chromatic router.
Triage scattered, nonlinear, or overloaded thinking into a structured stack of decisions, priorities, and handoffs. Use when the project feels ambiguous, when too many threads are running in parallel, when you cannot determine what to work on next, or when a session has accumulated more open questions than resolved ones.
Audit and redesign a repository's folder structure, root hygiene, scattered files, and naming conventions. Use when the repo root is cluttered, directories have grown organically without governance, files are misplaced across boundaries, or you need a clean canonical tree before onboarding agents or contributors.
Enforce numbered folder and subfolder naming standards across a repository. Audit whether directories follow an established numeric prefix convention (e.g. .01_, .02_, 01-, 1.) and produce a compliance report with remediation steps. Use when repo directories have grown without consistent naming, when onboarding contributors, or before a structural refactor.
Plan, audit, and standardize GitHub organization and repository governance — including personal-to-org repo transfers, branch protection, rulesets, team permissions, labels, repository settings, secrets review, and post-transfer validation. Use when moving repos into an organization, auditing existing permissions, generating settings checklists, or producing migration runbooks.
| name | agent-retrospective |
| description | Evaluate agent performance after a sprint, review, incident, build cycle, or multi-agent run. Use when the user asks what worked, what failed, which agents were effective, where delegation broke down, or how to improve future agent workflows. |
Turn agent workflow experience into measurable improvements for skills, prompts, policies, and review gates.
Produce a concise, auditable markdown artifact. Include:
## Agent Retrospective
**Scope:** ...
**Decision / recommendation:** ...
**Risk level:** low / medium / high / critical
### Findings
- ...
### Plan / matrix / record
| Item | Owner | Scope | Evidence | Gate / next action |
|---|---|---|---|---|
### Escalations
- ...
### Next actions
1. ...
Expected output focus: retrospective with scores, findings, root causes, improvement actions, owner, priority, next validation.
references/agent-retrospective-template.mdreferences/risk-tier-model.mdreferences/decision-rules.md