kmcp-dev-mcp-qa-e2e

Pass

Audited by Gen Agent Trust Hub on May 20, 2026

Risk Level: SAFE
Full Analysis
  • [PROMPT_INJECTION]: No patterns aimed at overriding agent behavior or bypassing safety guidelines were detected. The instructions follow a professional development workflow.- [DATA_EXFILTRATION]: The skill includes explicit instructions for secret redaction when generating raw JSON trace reports. Data storage is restricted to a local, gitignored directory (.local/), minimizing the risk of accidental exposure or exfiltration.- [COMMAND_EXECUTION]: Command execution is limited to standard local development scripts (e.g., npm run dev:test) and intended MCP tool calls. These operations are restricted to the local development environment.- [INDIRECT_PROMPT_INJECTION]: The skill possesses a surface for indirect injection as it processes MCP tool outputs and reads repository files. However, it mitigates risk through a phased workflow and mandatory secret redaction.
  • Ingestion points: call_mcp_tool responses and local repository file reads.
  • Boundary markers: Phased isolation (Phase 1 vs. Phase 2).
  • Capability inventory: File-system writes to .local/ and execution of npm test scripts.
  • Sanitization: Explicit requirement for raw JSON trace redaction.- [REMOTE_CODE_EXECUTION]: No patterns involving the download or execution of untrusted remote code were found.
Audit Metadata
Risk Level
SAFE
Analyzed
May 20, 2026, 07:40 AM
Security Audit — agent-trust-hub — kmcp-dev-mcp-qa-e2e