accint-frames
Pass
Audited by Gen Agent Trust Hub on Aug 11, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill exposes the agent to indirect prompt injection by instructing it to process data from an external "deliberation queue" without establishing safety boundaries or sanitization.
- Ingestion points: The agent retrieves data using the
acc framescommand, which includes a "typed hole" and "retrieved context" from an external queue (SKILL.md). - Boundary markers: The instructions contain no delimiters or specific directives to ignore instructions that might be embedded within the external context.
- Capability inventory: The agent uses the
acc_acttool to submit deliberate proposals back to the system, providing a write-back capability based on the external input. - Sanitization: The skill does not prescribe any validation, filtering, or escaping of the incoming queue data before it is processed by the agent.
Audit Metadata