RepoAI
Browse MCPs

Human Agent Trust Exploit Detection

100TrustedPre-approves tools

Identification of social engineering attempts and deceptive safety reassurances in chat.

What it does

Human Agent Trust Exploit Detection modifies your assistant's diagnostic layer to actively identify manipulative conversation patterns. When active, it scans for instances where the agent might be coerced into requesting sensitive credentials, providing false claims of security, or encouraging hazardous operational steps.

This skill is for auditing dialogue quality and safety compliance. By monitoring for specific behavioral indicators, it flags responses that misuse authority or pressure users into making poor security decisions. It ensures the assistant remains neutral and transparent rather than mirroring deceptive prompts.

When it earns its place

Reach for it when…

Install this when your agent interacts with users regarding system administration, security configurations, or handling sensitive organizational workflows. It is appropriate for environments where the agent provides actionable instructions that could result in data exposure or configuration errors if followed incorrectly.

Look elsewhere when…

Avoid this for basic informational retrieval or static creative writing tasks where the agent never requests or validates user data. It does not perform actual network security penetration testing or deep code analysis; it is strictly a semantic monitor for the assistant's own conversational output.

What to say to use it

You never invoke a skill directly — the assistant picks it up when your request matches what it covers. These are the kinds of things that do that.

  • Check if the agent is trying to extract my credentials
  • Monitor the conversation for deceptive security guarantees
  • Identify if the assistant is impersonating an administrator
  • Scan for risky operational advice in the chat history

Install it

Install for all your projects

Copy the skill directory into your personal skills folder. Claude picks it up on its own the next time your request matches what it does — there is nothing to enable.

bash
git clone --depth 1 --branch main https://github.com/Tencent/AI-Infra-Guard.git /tmp/AI-Infra-Guard
mkdir -p ~/.claude/skills
cp -r /tmp/AI-Infra-Guard/agent-scan/agent_scan/prompt/skills/human-agent-trust-exploit-detection ~/.claude/skills/human-agent-trust-exploit-detection

Install for one project only

Put it in the project's own skills folder instead, so it can be committed and shared with everyone working on that repository.

bash
cp -r /tmp/AI-Infra-Guard/agent-scan/agent_scan/prompt/skills/human-agent-trust-exploit-detection .claude/skills/human-agent-trust-exploit-detection

If a skill of the same name exists in both places, the personal one wins.

Installing on Claude.ai or through the API instead? The full guide covers every surface.

Pair it with an MCP server

Human Agent Trust Exploit Detection changes how the work is done. These servers change what Claude can reach while doing it.

How skills and MCP servers differ →