swarm
Pass
Audited by Gen Agent Trust Hub on Oct 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill defines a workflow for spawning multiple parallel workers and aggregating their terminal results into a final report, which creates a surface for instructions embedded in worker output to influence the agent.
- Ingestion points: The agent is instructed to read terminal results from workers in Phase C ('Aggregate').
- Boundary markers: There are no instructions to use delimiters or 'ignore' tags when processing the external output from these workers.
- Capability inventory: The skill utilizes the ability to spawn background subagents (
subagent_type: generalPurpose) and consolidate their results. - Sanitization: The instructions recommend against pasting raw dumps and instead favor a consolidated summary, which provides a natural layer of filtering but does not fully eliminate the risk of the agent following instructions contained within the worker's data.
Audit Metadata