skills/datadrivenconstruction/ddc_skills_for_ai_agents_in_construction/cwicr-unit-converter/Gen Agent Trust Hub
cwicr-unit-converter
Pass
Audited by Gen Agent Trust Hub on Sep 18, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process external datasets (CSV, Excel, JSON) which presents a potential surface for indirect prompt injection if the source files contain malicious instructions.
- Ingestion points: Input processing routines described in
instructions.mdand Python examples inSKILL.md(e.g.,pd.read_excel). - Boundary markers: No explicit delimiter or instruction-based boundary markers are defined to help the agent distinguish between data values and potential instructions within processed files.
- Capability inventory: Filesystem access for reading and writing data, and computational capabilities provided by the
pandasandnumpylibraries. - Sanitization: No specialized prompt-based filtering or sanitization for identifying adversarial content in the input data was identified.
Audit Metadata