agent-orchestration-multi-agent-optimize
Pass
Audited by Gen Agent Trust Hub on Oct 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input via arguments such as $TARGET and $PERFORMANCE_GOALS that are interpolated into agent instructions.
- Ingestion points: SKILL.md (Arguments Handling and Reference Workflows).
- Boundary markers: None present to delimit user-provided arguments from instructions.
- Capability inventory: The skill provides templates for Python-based profiling and orchestration, which an agent might implement (SKILL.md sections 1, 3, 5).
- Sanitization: No explicit input validation or escaping is defined for the provided arguments.
- [DYNAMIC_EXECUTION]: The skill provides static Python code templates for runtime agent orchestration, profiling, and cost optimization.
- Evidence: Code blocks in sections 1, 3, and 5 demonstrate creating classes and using concurrent.futures for managing agent execution logic.
Audit Metadata