product-research
Pass
Audited by Gen Agent Trust Hub on May 11, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill does not contain direct prompt injection or instructions to override agent safety. It does have an indirect prompt injection surface as it processes untrusted web data. 1. Ingestion points: External product pages are fetched using WebFetch and search snippets via WebSearch (SKILL.md). 2. Boundary markers: Absent; the agent is not explicitly instructed to ignore instructions within fetched content. 3. Capability inventory: The skill is authorized to update Google Sheets (mcp__google__sheets_values_update) and write to the local file system (Write, Edit) (SKILL.md). 4. Sanitization: None specified; fetched data is parsed and formatted directly into candidates.
- [DATA_EXFILTRATION]: No unauthorized data access or exfiltration patterns were found. Interactions with Google Sheets are consistent with the user's research workflow.
- [REMOTE_CODE_EXECUTION]: No patterns for downloading, installing, or executing remote code were identified.
- [COMMAND_EXECUTION]: The skill does not utilize bash or shell commands for sensitive system operations, privilege escalation, or persistence.
Audit Metadata