菜单
Multi-round ambiguity audit loop for a design document. Calls /design-evaluator (fresh-context agent) repeatedly, fixing gaps between rounds until CLEAN. Logs all findings, fixes, and decisions to .agent/audits/.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Universal entry point for any change to the project — new features, bug fixes, refactors, config changes, anything. Classifies the change; delegates contract-introducing changes (new feature, behavior change) to /plan-feature for ADR + verifier authoring; runs the dev-harness -> verify-suite loop until `sdd verify && verify[]` exits 0. Bug fixes and pure refactors against existing invariants skip ADR authoring per ADR-0078.
Universal entry point for any change to the project — new features, bug fixes, refactors, config changes, anything. Classifies the change; delegates contract-introducing changes (new feature, behavior change) to /plan-feature for ADR + verifier authoring; runs the dev-harness -> verify-suite loop until `sdd verify && verify[]` exits 0. Bug fixes and pure refactors against existing invariants skip ADR authoring per ADR-0078.
Universal entry point for any change to the project — new features, bug fixes, refactors, config changes, anything. Authors an ADR for the request, updates impacted specs (verified by spec-evaluator), and runs dev-harness -> implementation-evaluator loop until CLEAN. Handles spec backpressure via /plan-feature rules.
Given an ADR with an Invariant Delta block, invokes the invariant-compiler subagent to author per-invariant verifier code (test stubs, lint rules, schemas) for every Added entry and remove verifiers for every Withdrawn entry. Invokable inline by /plan-feature or standalone.
Multi-round ambiguity audit loop for a design document. Calls /design-evaluator (fresh-context agent) repeatedly, fixing gaps between rounds until CLEAN. Logs all findings, fixes, and decisions to .agent/audits/.
Record a bug with spec context so it's actionable for a future /feature-change session. Does not fix the bug — just documents it.
| name | design-audit |
| description | Multi-round ambiguity audit loop for a design document. Calls /design-evaluator (fresh-context agent) repeatedly, fixing gaps between rounds until CLEAN. Logs all findings, fixes, and decisions to .agent/audits/. |
| version | 1.0.0 |
| user_invocable | true |
| argument-hint | <path to design document> |
{{ include "skills/design-audit/SKILL.md" }}
