supervise
Pass
Audited by Gen Agent Trust Hub on Aug 24, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The supervisor role described in the skill integrates results, summaries, and report files produced by sub-agents into its own context. This creates a surface where a sub-agent processing untrusted data could return instructions that influence the supervisor's subsequent planning or dispatching actions.
- Ingestion points: Sub-agent verdicts, summaries (up to three lines), and full report files stored on disk (SKILL.md).
- Boundary markers: The skill instructions do not specify any delimiters or safety guidelines for processing or ignoring instructions embedded within sub-agent outputs.
- Capability inventory: The supervisor has capabilities to plan, dispatch fresh sub-agents (delegation), and maintain persistent state files for task resumption (SKILL.md).
- Sanitization: No sanitization, validation, or escaping of sub-agent content is mentioned before it is integrated into the supervisor's context.
Audit Metadata