digital-product-analysis
Pass
Audited by Gen Agent Trust Hub on Sep 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill possesses a documented surface for indirect prompt injection due to its core functionality of ingesting external data sources.
- Ingestion points: The skill reads the
REFERENCES.mdfile from the project root and is specifically instructed to follow links to local files (e.g., PDFs, spreadsheets) and access external URLs using theWebFetchtool. (Evidence inSKILL.mdunder Etapa 1). - Boundary markers: The instructions do not define clear delimiters or provide the agent with guidance to ignore potential instructions embedded within the user-provided data sources or external content.
- Capability inventory: The skill has access to
WebSearchandWebFetchtools for network reads and performs extensive file-system write operations to thedocs/directory. It does not appear to possess tools for arbitrary command execution or network exfiltration. - Sanitization: There is no evidence of validation or sanitization of the content fetched from external URLs or read from local files before it is processed by the agent.
Audit Metadata