swarm
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes terminal output from parallel workers, creating a surface for indirect instructions to be aggregated into the final report. \n
- Ingestion points: Phase C in SKILL.md specifies reading terminal results from N spawned workers. \n
- Boundary markers: The instructions do not specify the use of delimiters or warnings to ignore instructions within worker outputs. \n
- Capability inventory: The skill uses a spawn recipe from a pstack profile to execute tasks in cloud or local environments and consolidates findings into a report. \n
- Sanitization: The skill recommends summarizing rather than pasting raw worker dumps, but lacks formal sanitization or escaping of worker-provided data.
Audit Metadata