orchestrator
Pass
Audited by Gen Agent Trust Hub on Nov 13, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill acts as an orchestrator for complex user-defined tasks, creating a surface for potential indirect prompt injection where instructions embedded in processed data could influence sub-agent behavior.
- Ingestion points: The agent receives a complex objective from the user which is then decomposed into sub-tasks for workers in
SKILL.mdandSKILL.en.md. - Boundary markers: The protocol mandates a 'Work Contract' (Auftragsvertrag) for each worker, including explicit 'Positive Scope' and 'Negative Scope' definitions to restrict action and prevent instruction bleed.
- Capability inventory: The skill facilitates spawning workers capable of performing file edits, Git commits, research, and network operations as permitted by the environment.
- Sanitization: Implements a 'Review before commit' requirement and 'Readback at the destination' duty to verify all worker deliverables before integration into the main project.
Audit Metadata