dispatching-parallel-agents

Pass

Audited by Gen Agent Trust Hub on Mar 10, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill provides a structural framework and documentation for agent orchestration without implementing any dangerous functionality.
  • [DATA_EXPOSURE]: The file paths mentioned in examples (such as auth.ts, ~/.ssh/ is NOT present, but config/auth.ts is) are standard development placeholders and do not constitute an attempt to access sensitive system files.
  • [INDIRECT_PROMPT_INJECTION]: As a dispatching pattern, the skill inherently creates a surface for indirect prompt injection by generating instructions for sub-agents.
  • Ingestion points: Task descriptions and criteria are interpolated into sub-agent prompts via the 'Task Specification Template'.
  • Boundary markers: The skill uses clear Markdown headers (e.g., '### Objective', '### Scope') to separate instruction blocks, which helps reduce accidental obedience to data-embedded instructions.
  • Capability inventory: The skill itself does not perform file writes or network operations; it only suggests 'Verification Commands' like npm test for the agents to run.
  • Sanitization: No explicit sanitization logic is provided within the text templates.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 10, 2026, 05:57 PM