メニュー
Use when asking a question about the codebase, looking up a concept, or synthesizing an answer from the wiki + ref/ docs. Triggers: 'what do I know about X', 'look up Y', 'find in wiki', 'search docs', 'wiki says'. Read-only — defers writing to knowledge-capture.
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
| name | wiki-query |
| description | Use when asking a question about the codebase, looking up a concept, or synthesizing an answer from the wiki + ref/ docs. Triggers: 'what do I know about X', 'look up Y', 'find in wiki', 'search docs', 'wiki says'. Read-only — defers writing to knowledge-capture. |
| version | 1.0 |
Answer questions by synthesising the contents of docs/wiki/ (including ref/). Cite every claim with [Title](path) so the user can verify. Never write to disk. If the answer should be persisted, point the user to the knowledge-capture skill.
docs/wiki/core/00-system-index.md to enumerate all wiki docs with summaries.docs/wiki/ref/ref-index.md to enumerate all reference docs with topics.ref/ docs (e.g. tag writing → REF-Tag-Schedule.md).ref/ if topic-matched, rest from main wiki)."No relevant docs found in the wiki. The question may be out-of-scope, or you may want to add knowledge via the knowledge-capture skill first." Do NOT fabricate an answer from training data when no source is found.
[Title](relative-path) inline citations.In-conversation citations use project-root-relative paths:
[Article Title](docs/wiki/core/18-knowledge-capture.md)[Tag Schedule](docs/wiki/ref/REF-Tag-Schedule.md)Within a citation list at the bottom of the answer, prefer the same project-root-relative form.
Always close the response with:
"Want to persist this insight as a knowledge-capture decision? Use the
@knowledge-captureskill to add an entry."
This is a pointer, not an action. The user decides whether to write.
handleClick do?") — out of scope. Point to grep / codebase exploration, not the wiki.docs/logs/agent-changelog.md.@knowledge-capture or @wiki-lint skills.@knowledge-capture).Make sure to use this skill whenever the user mentions "pass the parcel", "parcel mode", "/parcel", "token saving planning", "multi-agent planning", "stateless execution", "clear context", "independent reviewer", or wants to run a highly token-efficient, robust design-and-execution pipeline where state is passed entirely within a .md plan in docs/plans/.
Activate this persona during Phase 7 and Phase 8 (Execution & QA Verification) of a parcel plan to execute codebase edits with absolute surgical precision, manage build/lint environments, and verify runtime stability.
Activate this persona during Phases 1, 2, and 3 (Scoping, Context Gathering, and User Clarification) of a parcel plan to lock down boundaries and eliminate ambiguity.
Activate this persona during code reviews, architectural planning, or specifically during Phase 6 (Senior Dev Hygiene Review) of a parcel plan to ruthlessly audit code quality, security, minimalism, and structural integrity.
Activate this persona during detailed architectural planning, or specifically during Phase 4 (Detailed Execution Plan) of a parcel plan to ruthlessly enforce the Simplicity Ladder, reject speculative scope, and produce surgically precise file-level blueprints.
Activate this persona during scoping, user flow design, or specifically during Phase 5 (Product Owner Review) of a parcel plan to ruthlessly smooth the product vision, user journey, and user experience by eliminating bloat and complexity.