skills.rest
blamejs
blamejsOFFICIAL
Explore all skills by this author →
WebsiteGitHub

mcp-agent-trust

Enumerate MCP trust boundary failures and assess compliance gaps.

Updated May 11, 2026
Search Tags:#ai security#prompt injection#supply chain security#agent trust#compliance gap analysis#mcp security#mcp rce
One-click install
npx skills add https://github.com/blamejs/exceptd-skills --skill mcp-agent-trust
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: mcp-agent-trust
Source: https://github.com/blamejs/exceptd-skills/tree/main/skills/mcp-agent-trust
Command: npx skills add https://github.com/blamejs/exceptd-skills --skill mcp-agent-trust

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill solves the critical gap where existing security and compliance frameworks (NIST 800-53, ISO 27001, SOC 2, EU AI Act) have no controls for Model Context Protocol (MCP) trust boundaries, leaving AI coding assistants exposed to local remote code execution, supply chain compromises, and indirect prompt injection attacks via malicious or compromised MCP servers.

Core Features & Use Cases

  • MCP Trust Boundary Failure Enumeration: Identifies unaddressed trust gaps including missing tool allowlisting, unsigned server manifests, absent bearer authentication, and zero-interaction RCE vectors.
  • Cross-Framework Gap Mapping: Flags where 10+ global compliance frameworks (NIST, ISO, SOC 2, SWIFT CSCF, EU NIS2/DORA, UK CAF, AU Essential 8) fail to cover MCP-specific risks, with explicit control-level gap declarations.
  • End-to-End Assessment Workflow: Provides a 5-step procedure to inventory installed MCP servers, verify package provenance, score trust posture, assess CVE exposure (including CVE-2026-30615 and CVE-2026-39884), and generate prioritized remediation actions.
  • Defensive Countermeasure Mapping: Aligns findings to D3FEND v1.3.0+ countermeasures for executable allowlisting, payload profiling, and authentication hardening.
  • Use Case: A security team using Cursor, Windsurf, or Claude Code can run this skill to audit all developer workstation MCP servers, remove unvetted tools, pin signed versions, and produce a compliance-ready gap report for their annual ISO 27001 audit.

Quick Start

Use the mcp-agent-trust skill to audit all installed MCP servers on your development workstations, flag unvetted tools with RCE or supply chain risk, and generate a compliance gap report for your security audit.

Frequently Asked Questions about mcp-agent-trust

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

FAQPage Schema
How do I audit MCP servers for remote code execution and supply chain risks?▼

Auditing MCP servers for RCE and supply chain risks involves enumerating installed servers, verifying package provenance, assessing CVE exposure, and scoring trust posture. This skill delivers a 5-step procedure to inventory tools, flag unsigned manifests, and generate remediation actions for AI coding assistants.

What is an MCP trust boundary failure in AI coding assistants?▼

An MCP trust boundary failure occurs when Model Context Protocol implementations lack tool allowlisting, unsigned server manifests, or bearer authentication. These unaddressed gaps expose AI coding assistants to local remote code execution, supply chain compromise, and indirect prompt injection attacks.

Does NIST 800-53 or ISO 27001 cover Model Context Protocol security risks?▼

Existing frameworks like NIST 800-53, ISO 27001, and SOC 2 have no controls for Model Context Protocol trust boundaries. This skill maps cross-framework gaps by flagging where 10+ global compliance frameworks fail to cover MCP-specific risks with explicit control-level declarations.

How do I generate a compliance gap report for MCP tools like Cursor and Claude Code?▼

Generating a compliance gap report for MCP tools requires auditing developer workstations, removing unvetted tools, and pinning signed versions. This skill produces compliance-ready gap declarations aligning findings to D3FEND countermeasures for ISO 27001, SOC 2, and EU AI Act audits.

Can I use this skill to check indirect prompt injection vulnerabilities in Gemini CLI?▼

Yes, you can use this skill to check indirect prompt injection vulnerabilities in Gemini CLI. It enumerates trust boundary failures across MCP-enabled AI tools including Cursor, Windsurf, Claude Code, and Gemini CLI, identifying zero-interaction RCE vectors and mapping defensive countermeasures.

What are the limitations of existing security frameworks for MCP supply chain compromise?▼

The limitation of existing security frameworks for MCP supply chain compromise is the total absence of controls for tool allowlisting and unsigned manifests. This skill addresses this by providing explicit control-level gap declarations for missing bearer authentication and zero-interaction RCE vectors.

Related Skills

Explore More Legal & Compliance Skills→
rwc9urwc9u

list-components

Discover installed Claude Code components and their provenance via a Ruby script.

Community
Intermediate
GRCEngClubGRCEngClub
367

finding-generator

Generate audit findings using the Condition-Criteria-Cause-Effect framework.

Official
Intermediate
hanabi-jpnhanabi-jpn
3

mac-sentinel

Harden macOS environments and detect threats for Claude Code.

Community
Advanced
pauljbernardpauljbernard
5

Red Team AI Security

Model advanced security threats targeting AI systems and development pipelines.

Community
Advanced