Pass
Audited by Gen Agent Trust Hub on May 7, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The Python script
scripts/pinterest_api.pychecks for the presence of thehttpxlibrary and automatically installs it from the official Python Package Index (PyPI) if missing.httpxis a well-known, reputable library for HTTP networking. - [COMMAND_EXECUTION]: To manage dependencies, the script executes the
pippackage manager usingsubprocess.check_call. This is a standard and safe implementation for a standalone utility script. - [PROMPT_INJECTION]: The skill processes untrusted content (pin descriptions and titles) from Pinterest search results, representing an indirect prompt injection surface.
- Ingestion points: Data is fetched via web scraping and API calls in
scripts/pinterest_api.pyand returned to the agent context. - Boundary markers: The instructions do not define explicit delimiters to isolate the scraped pin data from the agent's internal instructions.
- Capability inventory: The skill has the ability to perform network requests and execute local scripts.
- Sanitization: While descriptions are truncated, the content is not sanitized for potential embedded instructions before being passed to the agent.
- [DATA_EXFILTRATION]: All network requests are strictly scoped to official Pinterest infrastructure (
pinterest.com,api.pinterest.com) and its content delivery network (i.pinimg.com). No sensitive data access or third-party exfiltration was detected.
Audit Metadata