what-we-forgot
Pass
Audited by Gen Agent Trust Hub on May 14, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The instruction to "Review your instructions and extract all rules" is a form of system prompt extraction designed to reveal or refocus the agent on its underlying operational guidelines.
- [PROMPT_INJECTION]: The skill's design involves scanning untrusted conversation data for tasks, creating an indirect prompt injection surface.
- Ingestion points: Conversation history (referenced in SKILL.md)
- Boundary markers: Absent
- Capability inventory: General agent capabilities are used to perform the tasks found in the history
- Sanitization: Absent
Audit Metadata