skills.rest
msaleme
msalemeCOMMUNITY
Explore all skills by this author →
Website𝕏@mikesalemeGitHub

agent-security-harness

Validate AI agent security controls with protocol-level adversarial tests for MCP, A2A, L402, and x402 endpoints.

25|5|Updated Nov 15, 2025
Search Tags:#mcp#a2a#x402#protocol testing#agent security#l402#decision governance
One-click install
npx skills add https://github.com/msaleme/red-team-blue-team-agent-fabric --skill agent-security-harness
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: agent-security-harness
Source: https://github.com/msaleme/red-team-blue-team-agent-fabric/tree/main
Command: npx skills add https://github.com/msaleme/red-team-blue-team-agent-fabric --skill agent-security-harness

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires requests, geopy.

What problem does it solve?

Prevent authorized AI agents from being manipulated into unsafe or policy-violating behavior by testing the full protocol and decision-governance surface before you ship.

Core Features & Use Cases

  • Protocol integrity testing (MCP/A2A/L402/x402): probes wire-level message handling for spoofing, replay, downgrade, diversion, capability escalation, and tool registration poisoning.
  • Decision governance validation: checks whether agents should act at all under their authority using autonomy scoring, scope constraints, and unsafe-output defenses (including jailbreak and over-refusal behavior quality).
  • AIUC-1 aligned evidence output: produces structured JSON reports plus mappings for auditable pre-certification style workflows.
  • CI/CD and MCP server mode: runs as a CI gate and can expose the harness via MCP (defaulting to stdio to minimize network surface).

Example use case: you run the harness against your MCP server and confirm that malicious tool registration or capability escalation attempts are blocked, then attach the JSON evidence to your release gate.

Quick Start

Ask the AI assistant you’re using to run: "Run agent-security-harness MCP protocol tests in simulate mode first, then rerun against my MCP server URL if the simulate checks pass."

Frequently Asked Questions about agent-security-harness

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

FAQPage Schema
How do I test MCP protocol security for spoofing and replay attacks?▼

MCP protocol security testing for spoofing and replay attacks is executed by running an adversarial harness that probes wire-level message handling to detect manipulation. The harness runs protocol-level test suites against your endpoints and produces pass/fail results with timing metrics.

What is decision governance validation for AI agents?▼

Decision governance validation for AI agents checks whether systems act within their authority using autonomy scoring and scope constraints. It evaluates unsafe outputs including jailbreak behavior and over-refusal, ensuring agents do not violate policy before deployment.

Can I run agent security tests as a CI/CD gate?▼

Yes, you can run agent security tests as a CI/CD gate to block unsafe deployments. The harness validates protocol integrity and decision governance, then generates structured JSON evidence to satisfy pre-certification audit requirements for your release pipeline.

Does the security harness support A2A and x402 protocol testing?▼

Yes, the security harness supports A2A and x402 protocol testing alongside MCP and L402. It probes these endpoints for downgrade, diversion, capability escalation, and tool registration poisoning to ensure wire-protocol integrity before production deployment.

Do I need to provide my own credentials for agent security testing?▼

Yes, you need to provide operator-provided credentials scoped specifically for agent security testing. The harness enforces opt-in telemetry and requires operator-provided credentials to maintain strict authorization scope boundaries during protocol and governance validation.

What limitations exist when testing agent security in simulate mode?▼

Testing agent security in simulate mode runs checks without hitting live network surfaces, minimizing exposure. A limitation is that true wire-protocol integrity is only fully validated when subsequently rerunning the tests against your live MCP server URL.

Related Skills

Explore More Software Engineering Skills→
xkl2013xkl2013

response-compliance

Run OpenResponses compliance test suites against API endpoints.

Community
Intermediate
dotCMSdotCMS
949

cicd-diagnostics

Diagnose GitHub Actions CI/CD failures and generate issue reports.

Official
Advanced
BigsyBigsy
15

mcpmu

Multiplex multiple MCP servers behind a single stdio endpoint.

Community
Intermediate
barkleesandersbarkleesanders
5

tasks

Convert PRD markdown into prioritized executable task lists with acceptance criteria.

Community
Intermediate