crumb

Recommend a Crumb preset (ambient / video-research / mock) based on the user's setup intent. Trigger on "preset 추천", "추천해줘", "어떤 셋업", "which preset", "suggest a config", "혼자 돌릴까", or hints about installed tools / API keys. Output: recommended preset + reason + alternatives. Default recommendation = ambient (no --preset, entry host follows every actor). Crumb suggests only — user picks. Read-only. Do NOT trigger for game pitches (use `crumb` skill) or status checks (use `crumb-status` skill).

>-

Audit a feature/component for defects by cross-referencing with mature OSS patterns. Produce a prioritized matrix of defects + improvement proposals BEFORE implementing. Use when user says "결함 더 파고들어", "audit this", "find more defects", "improvement proposals first", "이 기능에서 뭐가 잘못됐는지 찾아".

>-

Auto-search and summarize latest AI/agent papers. Triggers: 'paper', '논문', 'arxiv', 'research', '연구', '최신 연구', '학회'.

CHANGELOG.md management rules — Keep a Changelog 1.1.0 + SemVer 2.0.0. Version releases, change logging, scope determination, format standards. Triggers on "changelog", "release", "version", "릴리스", "변경사항", "버전".

Scan the Obsidian wiki and automatically discover missing cross-references between pages. Use this skill when the user says "link my pages", "find missing links", "cross-reference", "connect my wiki", "add wikilinks", "what pages should be linked", or after any large ingestion to ensure new pages are woven into the existing knowledge graph. Also trigger when the user mentions "orphan pages" in the context of wanting to connect them, or says things like "my wiki feels disconnected" or "pages aren't linked well". This is a write-heavy skill — it actually modifies pages to add links, unlike wiki-lint which just reports issues.

Systematic web search, collection, analysis, and report generation on a topic. Triggers: 'research', '리서치', '조사해', '알아봐', '찾아봐', '트렌드', '동향'.

Branch strategy + PR rules. feature → develop → main merge flow (or feature → main when single-branch), Pre-PR Quality Gate (CI guardrails + docs-sync loop), Post-PR CI ratchet (gh pr checks --watch mandatory), Korean PRs, assignee settings. Triggers on "branch", "git", "pr", "merge", "커밋", "풀리퀘스트".

The foundational knowledge distillation pattern for building and maintaining an AI-powered Obsidian wiki. Based on Andrej Karpathy's LLM Wiki architecture. Use this skill whenever the user wants to understand the wiki pattern, set up a new knowledge base, or needs guidance on the three-layer architecture (raw sources → wiki → schema). Also use when discussing knowledge management strategy, wiki structure decisions, or how to organize distilled knowledge. This is the "theory" skill — other skills handle specific operations (ingesting, querying, linting).

Automated Git PR review — diff analysis + quality checks. Triggers: 'review', 'PR 리뷰', '코드 리뷰', 'diff', '풀리퀘스트 검토'.

Enforce consistent tagging across the Obsidian wiki using a controlled vocabulary. Use this skill when the user says "fix my tags", "normalize tags", "clean up tags", "tag audit", "what tags should I use", "tag taxonomy", or whenever you're creating or updating wiki pages and need to choose the right tags. Also trigger when the user asks about tag conventions, wants to add a new tag to the taxonomy, or says "my tags are a mess". Always consult this skill's taxonomy file before assigning tags to any wiki page.

Export the Obsidian wiki's knowledge graph to structured formats for use in external tools. Use this skill when the user says "export wiki", "export graph", "export to JSON", "export to Gephi", "export to Neo4j", "graphml", "visualize wiki", "knowledge graph export", or wants to use their wiki data in another tool. Outputs graph.json, graph.graphml, cypher.txt (Neo4j), and graph.html (interactive browser visualization) into a wiki-export/ directory at the vault root.

Ingest documents into the Obsidian wiki by distilling their knowledge into interconnected wiki pages. Use this skill whenever the user wants to add new sources to their wiki, process a document or directory, import articles, papers, or notes into their knowledge base, or says things like "add this to the wiki", "process these docs", "ingest this folder". Also triggers when the user drops a file and wants it incorporated into their existing knowledge base. Also handles raw mode: "process my drafts", "promote my raw pages", or any reference to the _raw/ staging directory.

Audit and maintain the health of the Obsidian wiki. Use this skill when the user wants to check their wiki for issues, find orphaned pages, detect contradictions, identify stale content, fix broken wikilinks, or perform general maintenance on their knowledge base. Also triggers on "clean up the wiki", "what needs fixing", "audit my notes", or "wiki health check".

Answer questions by searching the compiled Obsidian wiki. Use this skill when the user asks a question about their knowledge base, wants to find information across their wiki, asks "what do I know about X", "find everything related to Y", or wants synthesized answers with citations from their wiki pages. Also use when the user wants to explore connections between topics in their wiki. Works from any project. Includes an index-only fast mode triggered by "quick answer", "just scan", "don't read the pages", "fast lookup" — returns answers from page summaries and frontmatter without reading page bodies.

Archive existing wiki knowledge and rebuild from scratch, or restore from a previous archive. Use this skill when the user wants to start fresh, rebuild the wiki from all sources, archive current knowledge before a major change, or restore an older version. Triggers on "rebuild the wiki", "start over", "archive and rebuild", "restore from archive", "nuke and repave", "clean rebuild". Also use when the wiki has drifted too far from sources and incremental fixes won't cut it.

Initialize a new Obsidian wiki vault with the correct structure, special files, and configuration. Use this skill when the user wants to set up a new wiki from scratch, initialize the vault structure, create the .env file, or says things like "set up my wiki", "initialize obsidian", "create a new vault", "get started with the wiki". Also use when the user needs to reconfigure their existing vault or fix a broken setup.

Show the current state of the wiki — what's been ingested, what's pending, and the delta between sources and wiki content. Use this skill when the user asks "what's the status", "how much is ingested", "what's left to process", "show me the delta", "what changed since last ingest", "wiki dashboard", or wants an overview of their knowledge base health and completeness. Also use before deciding whether to append or rebuild. Includes an insights mode triggered by "wiki insights", "what's central", "show me the hubs", "central pages", "what's connected", "wiki structure" — analyzes the shape of the wiki itself to surface top hubs, cross-domain bridges, and orphan-adjacent pages.

Sync the current project's knowledge into the Obsidian wiki. Use this skill from any project when the user says "update wiki", "sync to wiki", "save this to my wiki", "update obsidian", or wants to distill what they've been working on into their knowledge base. This is the cross-project skill that lets you push knowledge from wherever you are into the vault.

LangGraph-style Mermaid diagram authoring guide. Applies the mermaid.live editor palette (Tailwind CSS-based). Use when generating architecture diagrams, StateGraph visualizations, or flowcharts. Triggers on the keywords "mermaid", "diagram", "flowchart", "topology", "graph".

Guide for creating effective skills. This skill should be used when users want to create a new skill (or update an existing skill) that extends Claude's capabilities with specialized knowledge, workflows, or tool integrations.

Diagnose Crumb routing faults against the F1-F7 fault matrix (adapter spawn fail / subprocess timeout / schema validation / qa.result missing / self-bias risk / infinite loop / env not propagated). Trigger on "왜 멈췄어?", "에러 났어?", "왜 안 돼?", "어디서 막혔어?", "session stuck", "debug", "fault", "이거 왜 이래?", or signs of trouble (no progress, repeated kinds, error events). Read-only. Returns table + suggested actions per detected fault.

Check whether the host environment can run Crumb — 3 host CLI authentication (Claude Code / Codex / Gemini), playwright availability, htmlhint fallback. Returns viable-preset table. Trigger on "환경 점검해줘", "doctor", "어떤 preset 가능?", "preset 가능한 거 알려줘", "is my setup ready?", "what can I run?", "환경 어때?", or before starting a fresh session in an unfamiliar environment. Read-only.

Export a Crumb session transcript to OpenTelemetry GenAI Semantic Conventions (otel-jsonl), Anthropic Console import format (anthropic-trace), or chrome://tracing (chrome-trace). Trigger on "OTel 로 변환", "datadog 으로", "Vertex 로", "trace export", "convert to OTel", "Anthropic Console 에 올려", "chrome tracing", or any request to ship the transcript to an external observability platform. Read-only.

Show current Crumb session status — last 10 signal events, latest D1-D6 scorecard with source-of-truth, total cost / cache hit / wall, stuck count. Trigger on "지금 어디까지 갔어?", "상황 어때?", "status", "session progress", "어떻게 진행 중?", "점수가 얼마야?", "scorecard 보여줘", or any "what's happening with my session" question. Read-only. Requires an active or recent session id (ULID). Do NOT trigger for new game pitches (use `crumb` skill) or preset questions (use `crumb-config` skill).

Recommend the user's next action in an active Crumb session — /approve, /veto, /redo, /pause, wait, or open summary. Branches on last event + verdict + audit + stuck_count. Trigger on "이제 뭐 하지?", "다음에 뭐 할까?", "what next", "next action", "다음 단계", "이거 끝난 거야?", "이거 어떻게 마무리?", or any "I'm not sure what to do" hesitation. Read-only. Output: primary recommendation + 1-3 alternatives. Do NOT trigger for status snapshots (use `crumb-status`) or fault diagnosis (use `crumb-debug`).

Open a previously generated Crumb game (single-file `game.html` or multi-file `index.html` + ES modules + service worker) in the browser. Trigger on "이전 게임 열어줘", "전에 만든 거 열어줘", "고양이 퍼즐 열어줘", "그 게임 다시 보자", "open the (cat / match-3 / merge / etc.) game", "open the previous game", "serve the multi-file game", "demo 열어줘", "결과물 열어줘", or any "show me what we built" phrasing. Multi-file games (`<script type="module">`, service worker, anything that needs CORS) MUST be served over http:// — `file://` protocol blocks ES module imports + service worker registration. Single-file games open with `file://` directly. The skill auto-detects which case applies, starts a local http server when needed, and opens the browser.

Explain a Crumb transcript event kind (39+ kinds: qa.result, judge.score, step.judge, user.intervene, handoff.rollback, etc.). Returns category, emitter, parent chain, payload shape, source-of-truth (D1-D6), and wiki spec ref. Trigger on "judge.score 가 뭐야?", "qa.result 는 뭔가요?", "explain <kind>", "이 kind 어디 emit?", "what is the X kind", "어떤 actor 가 emit해?", or any schema-vocabulary question. Read-only. Do NOT trigger for general "what is Crumb" or status questions.

Show or change Crumb per-actor model + effort + per-provider activation (claude-local / codex-local / gemini-cli-local). Defaults to all high-end models (claude-opus-4-7 / gpt-5.5-codex / gemini-3-1-pro) + effort=high. Gemini model IDs accept both dot ("gemini-3.1-pro") and dash ("gemini-3-1-pro") forms — internally normalized. Trigger on "verifier 모델을 X 로", "set builder model to Y", "effort 다 high 로", "effort all to low", "codex 비활성화", "disable gemini", "어떤 모델 쓰고 있어?", "모델 설정 보여줘", "show model config", "어떤 effort?", or any per-actor / per-provider tuning request. Read-only when no instruction given. Do NOT trigger for preset recommendation (use `crumb-config` skill) or environment readiness (use `crumb-doctor` skill).