skills/datadrivenconstruction/ddc_skills_for_ai_agents_in_construction/multi-agent-estimation/Gen Agent Trust Hub
multi-agent-estimation
Pass
Audited by Gen Agent Trust Hub on Sep 18, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill contains legitimate documentation, architectural diagrams, and code snippets for construction industry automation. All external references point to the author's official website or established open-source projects.
- [INDIRECT_PROMPT_INJECTION]: The skill handles untrusted construction data, creating a potential surface for indirect prompt injection. \n
- Ingestion points: The system ingests IFC models, PDF drawings, and project parameters (CSV/Excel/JSON) as described in SKILL.md and instructions.md. \n
- Boundary markers: Explicit delimiters are absent in the provided code snippets. \n
- Capability inventory: The skill uses filesystem and network permissions to read BIM data, process spreadsheets, and communicate with LLM providers. \n
- Sanitization: Present; instructions.md explicitly directs the agent to validate all inputs before processing and to follow industry standards, which significantly mitigates risk.
Audit Metadata