asb-carol-observations
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external sources and user-provided notes.
- Ingestion points: Public reviews, social posts, forum threads, and news articles retrieved through web searches, as well as 'team-mode' write-storm notes provided by the user.
- Boundary markers: The agent is instructed to present findings as 'candidate observations' that must be 'pressed and confirmed with the user' before being recorded.
- Capability inventory: The agent writes to the local file system (OBSERVATIONS.md) and is instructed to invoke other skills such as 'asb-rude-qa' and 'asb-carol-strengths'.
- Sanitization: The skill relies on manual user verification and clarification of every observation to ensure validity and safety of recorded content.
Audit Metadata