Skip to content

ActionGuard Protocol Inspector Agent

by lsendel · remote · no sign-in to list tools · v1.0.0 · published by GitHub user lsendel

Diff and poisoning detection for MCP servers: catch schema drift and prompt poisoning.

Changes · 2
CompareRegistry entry

2 tools: 2 change things.

Scanned Oct 21, 2026

1 check matched something in the tool definitions. No finding is not a safety guarantee. How the checks work.

claude mcp add --transport http actionguard https://actionguard.ai/mcp

Findings

Matches of fixed checks, with the text they matched.

1 warn
  • Warnactionguard_inspect

    The server marks "actionguard_inspect" read-only (readOnlyHint: true), and its own name, description or input give the class "write".

    Matched text

    annotations: {"readOnlyHint":true}; The name "actionguard_inspect" says "inspect": a word for reading.

    What this check looks for

Tools

Open a tool for its reasons and full description. Descriptions are the publisher's own words, shown as they are.

2 tools

Changes things 2Claude Code: ask
  • actionguard_check_driftVerify whether an MCP server's tools schema or descriptions have drifted from a pinned Dual-Hash.No findings

    Why it is in this group

    • Nothing in its name or description says what it does, so it is counted as a write.
    • The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.
    Description, as the server wrote it
    Verify whether an MCP server's tools schema or descriptions have drifted from a pinned Dual-Hash.

    Annotations the server declares

    • readOnlyHint: true
  • actionguard_inspectRun a 5-stage continuous reliability, security, and schema audit of an MCP server or AI agent endpoint.1 finding

    Why it is in this group

    • The name "actionguard_inspect" says "inspect": a word for reading.
    • The description's first verb is "run": a word for a change.
    • The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.

    Findings

    • Warn

      The server marks "actionguard_inspect" read-only (readOnlyHint: true), and its own name, description or input give the class "write".

      Matched text

      annotations: {"readOnlyHint":true}; The name "actionguard_inspect" says "inspect": a word for reading.

      What this check looks for
    Description, as the server wrote it
    Run a 5-stage continuous reliability, security, and schema audit of an MCP server or AI agent endpoint.

    Annotations the server declares

    • readOnlyHint: true

This is not a safety rating, and no finding is not a safety guarantee. The groups and findings come from fixed checks of the tool definitions the server returned on the date shown, not from its code or what it does when called. How the checks work.

Install

Pinned to version 1.0.0. Values in angle brackets are yours to fill in.

claude mcp add --transport http actionguard https://actionguard.ai/mcp

Run in a terminal.

Claude Code permissions

Allow the reads, ask before a change, deny the deletes. Paste into .claude/settings.json.

.claude/settings.json
{
  "permissions": {
    "allow": [],
    "ask": [
      "mcp__actionguard__actionguard_inspect",
      "mcp__actionguard__actionguard_check_drift"
    ],
    "deny": []
  }
}

Tools that only read are allowed, tools that change things ask first, tools that delete or cancel are denied. It assumes the server is installed as actionguard. Review it before you use it.