agenthub
Pass
Audited by Gen Agent Trust Hub on Nov 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill's primary function is to orchestrate multiple AI agents and aggregate their outputs. This creates a surface where instructions embedded in one agent's output could potentially influence the behavior of the orchestrator or downstream agents.\n
- Ingestion points: Agent outputs are ingested and stored in session files by
scripts/session_manager.pyand subsequently processed for ranking and merging byscripts/result_ranker.py.\n - Boundary markers: The skill architecture employs structured JSON outputs and explicit quality evaluation steps (
skills/eval.md) to validate outputs against defined criteria before they are passed to dependent agents.\n - Capability inventory: The system can perform filesystem writes (session state management) and trigger new agent executions based on the DAG definition.\n
- Sanitization: While the scripts enforce structured JSON formatting, they do not include specific filters for natural language prompt injection patterns within the content of the agent outputs.
Audit Metadata