health
Pass
Audited by Gen Agent Trust Hub on Apr 1, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes a comprehensive bash script and a python3 script to collect local project data, metrics, and configuration details for the health audit. This execution is limited to data collection within the local environment.
- [PROMPT_INJECTION]: The skill presents an indirect prompt injection surface due to its ingestion of third-party skill files and conversation logs. 1. Ingestion points: The Step 1 collection script reads conversation history (.jsonl) and external skill files (SKILL.md) into the agent context. 2. Boundary markers: The instructions for analysis subagents (Agent 1 and Agent 2) explicitly state to treat all pasted content as untrusted and to disregard any instructions found within that data. 3. Capability inventory: The skill uses bash and python3 for system analysis and configuration parsing. 4. Sanitization: The skill requires the primary agent to redact sensitive credentials, such as API keys and tokens, before sharing the collected data with analysis subagents.
Audit Metadata