# stand-out-of-my-sun/diogenes-agent-guardrails-mcp

Source: https://glimind.com/server/github%2Fstand-out-of-my-sun%2Fdiogenes-agent-guardrails-mcp
Registry id: `github/stand-out-of-my-sun/diogenes-agent-guardrails-mcp`
Status: unknown

## Tools (12)

- `audit_mcp_server_manifest`: Scans MCP server manifest for tool-name shadowing and prompt injection.
- `audit_outbound_egress_url`: Validates outbound URIs against cloud metadata and private IPs.
- `audit_response_factuality`: Cross-references claims against source passages to flag hallucinations.
- `detect_agent_action_loop`: Detects identical retry spins and runaway token burn.
- `detect_mcp_schema_rugpull`: Detects mid-session MCP tool schema tampering via structural diffs.
- `execute_python_sandbox`: Audits Python code for forbidden syscalls in a WebAssembly sandbox.
- `redact_sensitive_pii`: Redacts emails, SSNs, API keys, and crypto private keys.
- `sanitize_tool_output_stream`: Strips invisible Unicode and defuses indirect prompt overrides.
- `validate_mcp_sampling_request`: Blocks cross-server context snooping and token quota hijacking.
- `validate_sql_query_ast`: Validates SQL statements are strictly read-only SELECT clauses.
- `verify_a2a_agent_card`: Verifies A2A agent card JWS/DID signature and HTTPS enforcement.
- `verify_agent_command_safety`: Blocks prompt injection, destructive shell commands, and private-key exfiltration.

Measured by Glimind (https://glimind.com): MCP handshake + tool list only, no tool is ever called. CC BY 4.0.
