メニュー
Natural-language search across the Tool Intelligence Layer capability registry. Returns the top-5 capabilities that match a free-form query (e.g. 'I need to create a video'). Backed by semantic embeddings with keyword fallback.
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
Pipeline render layer and session log narrator for Mega Brain. Renders 24 canonical phase templates from PHASE-STREAM.jsonl, enforces truth-amplifier heuristics, and supports free-form mode for cross-phase signals.
Orquestracao multi-agente para clonagem cognitiva usando metodologia DNA Mental™ de 9 camadas.
Extrai conhecimento estruturado (frameworks, heurísticas, algoritmos, conceitos, metodologias, code snippets) de qualquer fonte: PDFs, sessões, transcrições, meetings, lives.
Extrai heurísticas operacionais de sessões de trabalho usando metodologia Pareto ao Cubo.
Use quando o operador pedir /força-total para um plano de orquestração global antes de executar: descobrir squads/capabilities, sintetizar DAG, paralelização, critical path, riscos FMEA, pre-mortem, handoff e next_action. SEMPRE despeja o plano completo (todos os agentes, workflows, tasks, ordem) e PARA para aprovação humana ANTES de avançar (Step 1 quality-gate). Triggers: força-total, forca-total, orquestração global, global orchestration plan, plan.yaml, plan.md, Mermaid DAG, critical path, capability discovery, pre-mortem. Produz plano global, DAG, resumo de 3 linhas, top risks e handoff. NOT for plano técnico local; use plan-architect para plano técnico especializado.
Orquestra cadeia completa de ingestao de transcricoes Fireflies (multi-account com dedup) -> /process-jarvis -> founder briefing cruzado com mission.md/north-star.yaml -> entrega via Z-API + Baileys (texto + audio ElevenLabs). Disparavel via terminal local OU via WhatsApp atraves do command_router.
| name | search-capabilities |
| description | Natural-language search across the Tool Intelligence Layer capability registry. Returns the top-5 capabilities that match a free-form query (e.g. 'I need to create a video'). Backed by semantic embeddings with keyword fallback. |
| version | 1.0.0 |
| owner_squad | infra-ops |
| megabrain_tier | Tier1 |
| context | fork |
| agent | general-purpose |
| user-invocable | true |
| argument-hint | [query string] |
| status | active |
Semantic + keyword tool discovery over agents/_registry/capability-registry.yaml.
Tool discovery: search_capabilities('your need') and you need to act on it.capability_id. Read the registry directly.enforcement_check / sync_capability_status.py instead.@devops + the tool-onboarder skill, not this one.python3 .claude/hooks/search_capabilities.py "your natural language query"
Add --json for machine-readable output, --top-k N to widen the result set (default 5).
A structured block:
Found N relevant capabilities:
[1] {capability_id} ({cosine score})
Title: ...
Description: ...
Business context: ...
Provider chain: ...
[2] ...
.data/capability-embeddings.npy (OpenAI text-embedding-3-large, 1536d).OPENAI_API_KEY is missing — substring scan against capability-keyword-index.json.| Path | Budget |
|---|---|
| Warm (embedding cache hit) | < 200 ms |
| Cold (cache miss, API call) | < 2 s |
| Keyword fallback | < 50 ms |
.claude/hooks/search_capabilities.py.claude/hooks/build_capability_embeddings.pyagents/_registry/capability-registry.yamldocs/stories/epic-til-foundation/STORY-TIL-19-tool-search-tool-defer-loading.mddocs/adr/ADR-TIL-001 (Python hooks boundary)