specify-high-precision-needs
Pass
Audited by Gen Agent Trust Hub on Oct 24, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill provides templates that involve processing potentially untrusted external content such as regulatory documents and source code.
- Ingestion points: The prompt templates in
SKILL.mdinclude placeholders for external data, notably[ATTACH DOCUMENT]and[PASTE SQL]. - Boundary markers: The instructions include a "HIGH-STAKES PRECISION DIRECTIVE" that forces the AI to use only provided facts and verbatim citations. While these are excellent defensive constraints, the templates do not explicitly recommend technical delimiters (like XML tags) to isolate the untrusted content.
- Capability inventory: The skill encourages the use of deterministic calculation tools (like Python) to review calculations or code.
- Sanitization: The skill relies on behavioral constraints ("Zero Extrapolation") rather than technical sanitization or escaping of the ingested data.
Audit Metadata