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 Mar 5, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: Fetches a Qdrant database snapshot from the official datadrivenconstruction GitHub repository.
- [EXTERNAL_DOWNLOADS]: Utilizes standard Python libraries including
qdrant-client,openai, andpandas. - [COMMAND_EXECUTION]: Provides instructions for running a Qdrant vector database container via Docker and downloading snapshots using
wget. - [PROMPT_INJECTION]: Identifies a surface for indirect prompt injection through the ingestion of external project data provided by the user in CSV, Excel, or JSON formats.
- Ingestion points: Project data files and user-provided parameters as defined in the input format instructions.
- Boundary markers: The skill lacks explicit boundary markers to delimit untrusted data from system instructions.
- Capability inventory: Includes filesystem access for reading data files and network access for API interactions with OpenAI and Qdrant.
- Sanitization: While basic input validation is mandated in the constraints, there is no specific sanitization of data content against embedded prompt injection attacks.
Audit Metadata