mcp-troubleshoot

Warn

Audited by Gen Agent Trust Hub on Feb 26, 2026

Risk Level: MEDIUMREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSCREDENTIALS_UNSAFE
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill instructs the user or agent to execute @gongrzhe/server-gmail-autoauth-mcp using npx. This command downloads and runs code from an external, untrusted source not associated with the skill author or a trusted vendor.\n- [EXTERNAL_DOWNLOADS]: The use of npx involves fetching packages from the NPM registry at runtime without version pinning, integrity checks, or source verification, representing an unverified dependency risk.\n- [CREDENTIALS_UNSAFE]: The documentation explicitly identifies the storage locations for sensitive Google OAuth keys and session tokens (~/.gmail-mcp/gcp-oauth.keys.json and ~/.gmail-mcp/credentials.json). Disclosing these paths to an AI agent increases the risk of data exposure or targeted exfiltration if the agent is subsequently compromised or misled.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Feb 26, 2026, 10:19 PM