skills/datadrivenconstruction/ddc_skills_for_ai_agents_in_construction/look-ahead-scheduler/Gen Agent Trust Hub
look-ahead-scheduler
Pass
Audited by Gen Agent Trust Hub on Mar 5, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: No security issues or malicious patterns were identified. The core logic consists of standard Python data structures and algorithms for construction scheduling and reporting.\n- [PROMPT_INJECTION]: The skill ingests user-provided project data in formats like CSV, Excel, and JSON. While this provides an indirect prompt injection surface, the instructions specifically mandate input validation and use the data for structured scheduling calculations rather than direct command execution.\n
- Ingestion points: instructions.md (User-provided project data files)\n
- Boundary markers: instructions.md (Instructions include 'Validate inputs before processing')\n
- Capability inventory: claw.json (filesystem permission for local data access)\n
- Sanitization: Noted in instructions as a requirement for the agent.\n- [EXTERNAL_DOWNLOADS]: The skill does not perform any network operations or external downloads. The SKILL.md file explicitly confirms it has no external dependencies.
Audit Metadata