decompose-mcp
por echology-io
AI & LLMsResumo
Decompose any text into classified semantic units — authority, risk, attention, entities.
Descrição
O que esta skill faz
Decomposes text or URLs into classified semantic units with authority levels, risk categories, attention scores, and entity extraction. No LLM or API calls involved. Runs locally with deterministic output.
Unlike LLM-based parsing, it produces identical output on repeated runs, costs nothing, works offline, and processes over 1,000 characters per millisecond.
Quando usá-la
- Pre-filtering a 200-page construction spec before sending to an LLM
- Extracting all OSHA and ASTM references from a safety policy
- Identifying mandatory vs. permissive clauses in a vendor contract
- Routing high-attention compliance sections to specialized review
- Building structured training datasets from raw regulatory documents
Exemplo de fluxo de trabalho
User asks: Decompose this contract and show me which clauses are safety-critical and mandatory
- 1
Call decompose_text with the contract text
- 2
Receive units array with authority and risk fields per clause
- 3
Filter units where authority is mandatory and risk is safety_critical
- 4
Rank filtered units by attention score descending
- 5
Return the prioritized list with heading_path context
A ranked list of mandatory safety-critical clauses with their document locations and attention scores
Instalação
clawhub install echology-io/decompose-mcpExecute o comando da CLI ou cole o prompt no Claude Code, Cursor, Codex ou no agente de código que você usa.