choose-your-orchestrator
Pass
Audited by Gen Agent Trust Hub on Nov 13, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill is designed to improve agent security by establishing clear boundaries and constraints for multi-agent tasks. It explicitly forbids the agent from expanding its authority or bypassing user approval for sensitive actions.
- [COMMAND_EXECUTION]: The skill utilizes local CLI tools
clutchandusmc. These are employed for routing suggestions and generating session notes. The instructions provide safety guidelines, such as treating tool output as a suggestion rather than an absolute authority and ensuring that session notes remain privacy-neutral. - [INDIRECT_PROMPT_INJECTION]: The skill processes user-provided task descriptions to recommend orchestration strategies.
- Ingestion points: The agent interpolates task descriptions into the
clutchtool and uses the resulting JSON to build recommendations. - Boundary markers: The 'Session Orchestration Contract' format provides a structured boundary for the final output.
- Capability inventory: The skill defines the configuration (parallelism, write boundaries, model slots) for subsequent agent actions.
- Sanitization: The instructions explicitly warn against storing raw prompts, secrets, or private content in external logs (USMC).
Audit Metadata