skills/datadrivenconstruction/ddc_skills_for_ai_agents_in_construction/semantic-search-cwicr/Gen Agent Trust Hub
semantic-search-cwicr
Pass
Audited by Gen Agent Trust Hub on Sep 18, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted project data (CSV, Excel, JSON) and file paths to perform semantic searches and cost estimations. While the instructions mandate input validation, malicious content within these files could potentially influence the agent's output or reasoning during the estimation process.
- Ingestion points: Project data and file parameters defined in instructions.md.
- Boundary markers: Not explicitly defined for data processing.
- Capability inventory: Vector search via qdrant-client and data manipulation via pandas.
- Sanitization: The skill includes a general instruction to validate inputs before processing.
- [EXTERNAL_DOWNLOADS]: The skill requires the installation of standard Python packages (pandas, qdrant-client, sentence-transformers) and downloads the BAAI/bge-m3 embedding model from Hugging Face. These are well-known and reputable sources in the AI/data science community.
Audit Metadata