cro-optimization
Pass
Audited by Gen Agent Trust Hub on Feb 15, 2026
Risk Level: LOWPROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- Prompt Injection (LOW): The skill processes untrusted external content, creating a vulnerability to indirect prompt injection where instructions in the analyzed page could influence the agent.
- Ingestion points: Landing page content is retrieved via the
WebFetchtool or accepted via$ARGUMENTS(File: SKILL.md, Step 3). - Boundary markers: Absent. No delimiters or 'ignore instructions' directives are used for the external data.
- Capability inventory: The skill uses
ReadandWebFetchbut does not include tools for arbitrary command execution or file system writing. - Sanitization: There is no evidence of sanitization or filtering of the fetched data before analysis.
- Data Exfiltration (LOW): The skill uses
WebFetchto access arbitrary, non-whitelisted domains to retrieve content for analysis (File: SKILL.md, Step 3).
Audit Metadata