skills.rest
unclenate
unclenateCOMMUNITY
Explore all skills by this author →
Website𝕏@unclenateGitHub

harness-mcp

Enforce governance-compliant MCP server behavior with tool tier mappings and threat documentation.

3|1|Updated Apr 5, 2026
Search Tags:#mcp#model context protocol#prompt injection defense#trust tiers#tool governance#oauth authorization#capability negotiation
One-click install
npx skills add https://github.com/unclenate/auto-harness --skill harness-mcp-unclenate
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: harness-mcp
Source: https://github.com/unclenate/auto-harness/tree/main/platform/skills/harness-mcp
Command: npx skills add https://github.com/unclenate/auto-harness --skill harness-mcp-unclenate

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill ensures that an AI-assisted project that produces an MCP (Model Context Protocol) server maintains the same governance, trust boundaries, and safety posture that would otherwise be lost when tools and runtime agents communicate over MCP.

Core Features & Use Cases

  • MCP producer-side governance: Defines how the harness tier model must be applied when your server exposes MCP tools to consumers.
  • Explicit tool tier mapping & review gates: Requires a one-line rationale per tool tier in docs/mcp/tool-registry.md to prevent placeholder defaults.
  • Threat surface coverage for MCP-specific attacks: Documents prompt-injection, capability drift, token handling, OAuth proxy confused-deputy, SSRF, and session/auth discipline requirements.
  • Honest capability negotiation checks: Enforces consistency between initialize declarations and runtime tool/capability behavior.
  • Transport and auth posture: Sets expectations for stdio vs HTTP transports, including bearer-token enforcement and refusal to forward inbound tokens.
  • Clear boundaries between sibling skills: Keeps producer governance distinct from harness-tools (consumer side) and harness-governance (trust tiers/lifecycle).

Quick Start

Use the harness-mcp skill when your project ships an MCP server that exposes tools to MCP clients, and tell your agent: “Review the MCP server changes and ensure tool registry tier mappings with rationales, capability negotiation consistency, prompt-injection test coverage, and transport/auth rules are all satisfied.”

Frequently Asked Questions about harness-mcp

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I secure an MCP server against prompt injection attacks?▼

To defend an MCP server against prompt injection, you must document threat surfaces like capability drift and token handling, and enforce consistency between initialize declarations and runtime tool behavior to prevent unauthorized actions.

What is MCP producer-side governance and when do I need it?▼

MCP producer-side governance applies trust tier models to tools your server exposes to consumers. You need it when your project ships an MCP server, ensuring safety boundaries are maintained over the protocol.

How do I map MCP tools to consumer trust tiers?▼

You map MCP tools to consumer trust tiers by providing a one-line rationale for each tool tier in your tool registry, preventing placeholder defaults and ensuring explicit boundaries are documented for review.

Does MCP capability negotiation require runtime consistency checks?▼

Yes, MCP capability negotiation requires runtime consistency checks to enforce honesty between initialize declarations and actual tool behavior, preventing capability drift and ensuring secure client-server interactions.

What transport and OAuth authorization rules apply to MCP servers?▼

MCP servers must enforce transport-specific auth discipline, including bearer-token enforcement for HTTP transports and refusing to forward inbound tokens, to prevent OAuth proxy confused-deputy and SSRF vulnerabilities.

Can I use MCP governance for consumer-side tool evaluation?▼

No, MCP producer-side governance is distinct from consumer-side tool evaluation. It specifically governs servers exposing tools to clients, keeping producer boundaries separate from consumer trust tier lifecycle management.

Related Skills

Explore More Legal & Compliance Skills→
Factory-AIFactory-AI
101

threat-model-generation

Generate a STRIDE-based threat model with threat-model.md and security-config.json.

Official
Advanced
corticalstackcorticalstack
1

mcp-developer

Implement MCP server and client integrations over JSON-RPC 2.0.

Community
Advanced
unclenateunclenate
3

harness-governance

Enforces trust-tier rules, companion document requirements, and lifecycle readiness checks for governed projects.

Community
Advanced
dreadnodedreadnode
11

mcp-auth-exploitation

Test MCP authorization flows and OIDC Dynamic Client Registration for SSRF vulnerabilities.

Official
Advanced