audit-support
Pass
Audited by Gen Agent Trust Hub on Oct 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external evidence, search results, and scraped data, creating a potential surface for indirect prompt injection where malicious instructions could be embedded in external sources.
- Ingestion points: Data retrieved via
sandbase_runfor external evidence, search, scraping, and enrichment as described inSKILL.mdandreferences/sandbase-api-map.md. - Boundary markers: While explicit delimiters are not defined, the skill mandates preserving authorization boundaries and verifying outputs against source material.
- Capability inventory: The skill utilizes the
sandbase_runtool to interact with external capabilities and generative models. - Sanitization: Instructions focus on independent verification of calculations and evidence of validation, though they do not specify technical escaping or filtering of external content.
Audit Metadata