ma-playbook
Pass
Audited by Gen Agent Trust Hub on Oct 27, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The
due_diligence_tracker.pyutility allows for the storage of user-provided strings (items and notes) in a local JSON file. This data is retrieved and processed by the agent to generate reports, creating a surface where malicious instructions embedded in the tracking data could attempt to influence the agent's behavior. - Ingestion points: CLI arguments
--itemand--notesinscripts/due_diligence_tracker.pyare saved todd_data.json. - Boundary markers: The script and instructions do not implement explicit delimiters or boundary markers to isolate untrusted data when it is presented back to the agent.
- Capability inventory: The skill includes local file read/write capabilities for state management (
dd_data.json) but lacks high-risk capabilities such as network access, arbitrary subprocess execution, or system-level writes. - Sanitization: No validation, escaping, or filtering of the input strings is performed before they are stored or re-injected into the agent's context.
Audit Metadata