Decompose Mcp

Decompose any text into classified semantic units — authority, risk, attention, entities. No LLM. Deterministic.

Security Scan

The skill's declared purpose and requirements mostly line up (local text decomposition + optional URL fetch), but several runtime claims (SSRF protections, fully local behavior) rely on external, install-time code (a PyPI package) that this manifest can't verify — so proceed with caution and audit the package before trusting network-enabled use.

Purpose & Capability

Name/description, declared binaries (python3), and the provided tools (decompose_text, decompose_url) are coherent: a Python-based decomposition tool reasonably needs Python and may need network for URL fetching. No unrelated credentials or unusual binaries are requested.

Instruction Scope

SKILL.md stays within the stated purpose: it shows how to install the package, configure an MCP server, and use text- vs URL-based tools. However it instructs running a local MCP server (python -m decompose --serve) which will start process(es) on the host and may accept local network connections. The doc claims SSRF protection and wholly-local text processing for decompose_text — those are implementation details referenced but cannot be validated from the instruction-only package. The agent instructions do not ask for unrelated files or secrets.

Install Mechanism

Install uses pip (uv) to fetch a PyPI package (decompose-mcp). Pulling from PyPI is common but executes third-party code on the host; this is a moderate-risk install mechanism because the actual package contents are not included here. The SKILL.md links to a GitHub repo (good for audit), but the skill bundle itself provides no code to verify the claims.

Credentials

No environment variables or credentials are requested. The only declared permission is 'network', which matches the documented decompose_url functionality and is appropriately scoped in the manifest. No other secrets or unrelated service tokens are required.

Persistence & Privilege

always:false (normal). The skill suggests adding an MCP server entry to your OpenClaw config (a user action), but does not request forced inclusion or modification of other skills. Running a local service is normal for MCP patterns but increases runtime footprint.

What to consider before installing

This skill appears to do what it says, but the runtime behavior depends on a PyPI package you must install. Before installing or enabling network-enabled features: 1) Inspect the referenced GitHub repo and the exact PyPI release (verify source-to-package integrity and review code, especially URL fetch and SSRF logic). 2) Prefer using decompose_text only if you do not need URL fetching (that avoids outbound requests). 3) Run the package in a sandboxed environment or ephemeral container first to observe network behavior. 4) Pin to a specific package version and review its PyPI release metadata. 5) If you will enable the MCP server, restrict its network exposure (bind to localhost only) and review firewall rules. If you cannot audit the package source or verify the SSRF protections, treat network-enabled usage as higher risk.

Decompose

Decompose any text or URL into classified semantic units. Each unit gets authority level, risk category, attention score, entity extraction, and irreducibility flags. No LLM required. Deterministic. Runs locally.

Setup

1. Install

pip install decompose-mcp

2. Configure MCP Server

Add to your OpenClaw MCP config:

{
  "mcpServers": {
    "decompose": {
      "command": "python3",
      "args": ["-m", "decompose", "--serve"]
    }
  }
}

3. Verify

python3 -m decompose --text "The contractor shall provide all materials per ASTM C150-20."

Available Tools

decompose_text

Decompose any text into classified semantic units.

Parameters:

• text (required) — The text to decompose
• compact (optional, default: false) — Omit zero-value fields for smaller output
• chunk_size (optional, default: 2000) — Max characters per unit

Example prompt: "Decompose this spec and tell me which sections are mandatory"

Returns: JSON with units array. Each unit contains:

• authority — mandatory, prohibitive, directive, permissive, conditional, informational
• risk — safety_critical, security, compliance, financial, contractual, advisory, informational
• attention — 0.0 to 10.0 priority score
• actionable — whether someone needs to act on this
• irreducible — whether content must be preserved verbatim
• entities — referenced standards and codes (ASTM, ASCE, IBC, OSHA, etc.)
• dates — extracted date references
• financial — extracted dollar amounts and percentages
• heading_path — document structure hierarchy

decompose_url

Fetch a URL and decompose its content. Handles HTML, Markdown, and plain text.

Parameters:

• url (required) — URL to fetch and decompose
• compact (optional, default: false) — Omit zero-value fields

Example prompt: "Decompose https://spec.example.com/transport and show me the security requirements"

What It Detects

• Authority levels — RFC 2119 keywords: "shall" = mandatory, "should" = directive, "may" = permissive
• Risk categories — safety-critical, security, compliance, financial, contractual
• Attention scoring — authority weight x risk multiplier, 0-10 scale
• Standards references — ASTM, ASCE, IBC, OSHA, ACI, AISC, AWS, ISO, EN
• Financial values — dollar amounts, percentages, retainage, liquidated damages
• Dates — deadlines, milestones, notice periods
• Irreducibility — legal mandates, threshold values, formulas that cannot be paraphrased

Use Cases

• Pre-process documents before sending to your LLM — save 60-80% of context window
• Classify specs, contracts, policies, regulations by obligation level
• Extract standards references and compliance requirements
• Route high-attention content to specialized analysis chains
• Build structured training data from raw documents

Performance

• ~14ms average per document on Apple Silicon
• 1,000+ chars/ms throughput
• Zero API calls, zero cost, works offline
• Deterministic — same input always produces same output

Security & Trust

Text classification is fully local. The decompose_text tool performs all processing in-process with no network I/O. No data leaves your machine.

URL fetching performs outbound HTTP requests. The decompose_url tool fetches the target URL, which necessarily involves network I/O to the specified host. This is why the skill declares the network permission in claw.json. If you do not need URL fetching, you can use decompose_text exclusively with no network access required.

SSRF protection. URL fetching blocks private/internal IP ranges before connecting: 0.0.0.0/8, 10.0.0.0/8, 100.64.0.0/10, 127.0.0.0/8, 169.254.0.0/16, 172.16.0.0/12, 192.168.0.0/16, ::1/128, fc00::/7, fe80::/10. The implementation resolves the hostname via DNS before connecting and checks all returned addresses against the blocklist. See src/decompose/mcp_server.py lines 19-49.

No API keys or credentials required. No external services are contacted except when using decompose_url to fetch user-specified URLs.

Source code is fully auditable. The complete source is published at github.com/echology-io/decompose. The PyPI package is built from this repo via GitHub Actions (publish.yml) using PyPI Trusted Publishers (OIDC), so the published artifact is traceable to a specific commit.

Resources

• Source Code (GitHub) — full source, auditable
• PyPI — published via Trusted Publishers
• Documentation
• Blog: When Regex Beats an LLM
• Blog: Why Your Agent Needs a Cognitive Primitive