agent-delegation
Pass
Audited by Gen Agent Trust Hub on Oct 31, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill's supervisor-worker architecture involves ingesting artifacts and progress updates from specialized worker agents.
- Ingestion points: Data enters through contracts/schemas/a2a-artifact.json and a2a-task-progress.json.
- Boundary markers: Instructions explicitly mitigate goal hijacking by treating worker content as data and promoting isolated contexts.
- Capability inventory: The supervisor role has access to powerful tools including execute_command, write_file, and run_tests.
- Sanitization: The skill enforces schema validation for all returned artifacts, though it does not specify sanitization for natural language fields contained within the data.
Audit Metadata