multi-service-debug

Pass

Audited by Gen Agent Trust Hub on Feb 25, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill provides numerous bash workflows utilizing gcloud, vercel, and curl to inspect service configurations, check IAM policies, and fetch logs. These tools represent a powerful interface for interacting with production environments during troubleshooting sessions.
  • [EXTERNAL_DOWNLOADS]: The documentation recommends the installation of third-party debugging utilities via brew and npm. Additionally, it includes a command to execute wscat using npx, which involves fetching and running code from the npm registry at runtime.
  • [PROMPT_INJECTION]: The skill facilitates the ingestion of external log data which creates a surface for indirect prompt injection. An attacker could craft specific log messages designed to influence the agent's reasoning when it processes those logs.
  • Ingestion points: Service logs fetched via gcloud logging read and vercel logs in SKILL.md.
  • Boundary markers: Absent. The instructions do not specify delimiters to isolate untrusted log content from the agent's core instructions.
  • Capability inventory: Access to cloud management CLIs (gcloud, vercel), network request utilities (curl), and WebSocket debuggers (websocat, wscat).
  • Sanitization: Absent. There are no steps provided to sanitize or validate the content of the logs before they are analyzed by the agent.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 25, 2026, 03:31 PM