메뉴
기본적으로 소스를 먼저 확인하는 Prompt가 선택됩니다. 직접 명령으로 전환하거나 로컬 사본을 다운로드할 수도 있습니다.
설치 여부를 결정하기 전에 SKILL.md와 SkillsMP에 표시된 보조 파일을 읽어 보세요.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
직접 명령은 검토 Prompt를 거치지 않습니다. 실행하기 전에 소스를 확인하세요.
npx skills add https://github.com/SeemSeam/agent-roles-spec --skill requirement-reanalyze명령은 한 줄로 유지됩니다. 복사하기 전에 가로로 스크롤해 전체 내용을 확인하세요.
로컬 사본을 원하시나요? SkillsMP에서 현재 제공할 수 있는 파일을 다운로드하세요.
Review a declared code change or code scope with evidence-backed findings and five independent quality scores. Use for pull requests, diffs, commits, patches, file sets, modules, bug fixes, refactors, and implementation-review requests.
Diagnose and recover a CCB agent whose provider context is corrupted, stale, overloaded, or semantically off-track when the user wants more than a plain clear: clear the agent only after safety checks, reconstruct the active task context from CCB lineage/runtime evidence, and resume or resubmit the interrupted work with a compact handoff. Use for requests like "clear this agent then restore its task", "context is broken but continue the current job", "repair talk1/worker context and inherit current state", or "after clear, recover the task".
Diagnose and recover CCB communication and reply delivery stalls. Use when a user reports a missing CCB_REPLY, stuck ask, agent stuck busy/delivering, queued work behind an active job, cancelled/incomplete reply, empty artifact, chain not continuing, rejected active-turn followup, duplicate retry after success, or a CCB mailbox/communication backend that appears stuck.
SOC 직업 분류 기준
SKILL.md 표시 중
| name | requirement-reanalyze |
| description | 重新分析 Requirement 原文和已生成人读设计/任务文档，修正理解漂移。 |
| metadata | {"short-description":"Requirement 重新分析"} |
requirement-reanalyze 用于当用户、Claude 或 Codex 发现需求理解可能跑偏，或需求内容本身被用户变更 / 扩张时，重新进入需求分析。它不只是刷新摘要，而是重新质疑 framing、歧义、用户必问项和已有设计 / dev_task 文档。
本 skill 同时处理两类场景：① 理解漂移（对同一内容的旧判断纠偏）；② 需求内容变更 / 新增（以新内容重推范围）。两类都遵守范围非锚定：以用户最新完整原文为准，旧设计 / dev_task 只作历史证据、不绑定新增内容；必须先核验当前代码 / 事实，再判断新范围是否仍命中既有 API / schema / 依赖 / 成本 / 合规约束，并明确哪些旧结论保留、哪些被推翻、哪些需进入 technical_design 或 task_breakdown。禁止用旧边界给新增内容设准入门槛，也禁止凭未核验假设放大 / 缩小范围。
主要进入：
| 节点 | Manifest |
|---|---|
| 需求分析 | references/kernel/nodes/requirement_analysis.node.md |
可能继续：
| 场景 | 下一节点 |
|---|---|
| 需求重新澄清后需要改方案 | references/kernel/nodes/technical_design.node.md |
| 只影响拆分 | references/kernel/nodes/task_breakdown.node.md |
/ccb:requirement-reanalyze requirement_id=<id>
/ccb:requirement-reanalyze docs/02_需求设计/foo.md
如果已有设计 / dev_task 与用户原文冲突，以用户原文和用户最新决定为更高优先级。
启动时先读 docs/00_项目总览.md、docs/00_文档地图.md 和 docs/.ccb/docs-structure-contract.yaml。直接读取经 resolver 定位的 requirement 文档、关联 technical_design / dev_task 文档、breakdown draft 和 EventJournal。修订后的分析写回 requirement markdown，并在 frontmatter 写入：
analysis_input_hash：本次分析对应的 title + 需求描述 hash。analysis_applied_at：本次主动应用分析的 ISO8601 时间戳。Console 如存在，只通过 file watcher 解析这些 frontmatter 字段并投影到 DB；本 skill 不调用 Console HTTP API。
修订分析必须写入 analysisFile JSON 后调用 applyRequirementAnalysis({ projectRoot, requirementId, analysisFile })。analysisFile 必填非空字符串字段：claudeInterpretation、ambiguities、fidelityDiff；可选字符串字段 bodyMarkdown 用于按需求模板主体产出二~十三章（复杂度自适应，用不上的章节可删）。bodyMarkdown 不得包含以下二级标题：需求描述、原话（verbatim）、原话、verbatim、Claude 解读、Claude 解读（可选）、歧义点、歧义点（可选）、保真差异、保真差异（可选）。提供 bodyMarkdown 时，lib 会以 O2 语义整段替换「原话 section 后 ~ Claude 解读 section 前」的主体区；未提供时保持旧三锚点写入路径。严禁绕过 lib 直接写 requirement markdown。
必须执行需求分析节点的 sc 和协商要求。特别要求 Codex 对比“旧理解 vs 新原文”，指出是否存在 semantic drift。
输出漂移点、保留不变的结论、已在终端问到答案并闭环记录的用户拍板项、写入路径和后续节点建议；若仍有命中项未获答复，先停在当前终端等待答案，不落档。