dispatching-parallel-agents
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill describes a parallel processing pattern that involves ingesting content from sub-agents, which acts as an untrusted data source.
- Ingestion points: The instructions in SKILL.md under 'Review and Integrate' direct the agent to 'Read each summary' returned by sub-agents.
- Boundary markers: The skill lacks explicit boundary markers or instructions to treat the sub-agent output as untrusted data that should not be interpreted as commands.
- Capability inventory: The skill references capabilities to spawn tasks (e.g., the 'Task()' function) and modify source code to fix bugs (SKILL.md sections 'Dispatch in Parallel' and 'Agent Prompt Structure').
- Sanitization: There are no described sanitization or validation steps for the summaries before the parent agent integrates the changes.
Audit Metadata