audit-agentic-mcp
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill is a legitimate development and security auditing tool for the Model Context Protocol (MCP) ecosystem.
- [COMMAND_EXECUTION]: The provided bash scripts (
audit-mcp-server.shandmeasure-context-budget.sh) utilize standard read-only commands such asgrep,rg,awk, andfindto perform static analysis on local codebase directories. These operations are intended strictly for reporting on code structure and token usage and do not involve destructive or network-active commands. - [PROMPT_INJECTION]: Static analysis warnings regarding prompt injection keywords are false positives. Phrases like 'ignore previous instructions' appear exclusively within the security documentation as examples of malicious patterns that server authors should detect and mitigate, rather than as instructions for the agent to bypass its own constraints.
- [EXTERNAL_DOWNLOADS]: The skill contains references to numerous authoritative and well-known resources, including official MCP documentation, platform specifications from major AI vendors (Anthropic, OpenAI, Google), and academic research hosted on arXiv. All links are informative and contextually appropriate for an auditing framework.
Audit Metadata