asb-carol-observations

Pass

Audited by Gen Agent Trust Hub on Aug 17, 2026

Risk Level: SAFE
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill includes instructions to ingest untrusted data from public reviews, social posts, and press articles through web searches. It effectively mitigates potential injection risks by placing this data in a dedicated 'External Research' section and requiring the user to explicitly confirm, correct, or reject these findings before they are promoted to the final observation list.
  • [COMMAND_EXECUTION]: The skill facilitates the invocation of other specific AI agent skills from the same author, such as asb-rude-qa and asb-carol-strengths. These invocations are for structured quality checks and subsequent analysis steps, representing normal tool-chaining behavior rather than arbitrary command execution.
  • [DATA_EXPOSURE_AND_EXFILTRATION]: The skill manages session state by writing to a local file named OBSERVATIONS.md in a user-defined directory. It does not attempt to access sensitive system files or credentials, and its network operations are limited to gathering public marketing and review information via standard web search tools.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 17, 2026, 03:31 PM
Security Audit — agent-trust-hub — asb-carol-observations