bim-clash-detection

Pass

Audited by Gen Agent Trust Hub on Mar 5, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection through the processing of untrusted BIM model data.
  • Ingestion points: Data is loaded from external files via the load_elements method in SKILL.md which converts element properties into strings for analysis and reporting.
  • Boundary markers: The skill lacks boundary markers or specific instructions to the agent to isolate element metadata from its own operational logic.
  • Capability inventory: The skill is granted filesystem permissions in claw.json, which allows it to read and write files based on user/data inputs.
  • Sanitization: Component attributes such as names, categories, and level descriptions are processed as raw strings without validation or filtering for potential instructional content.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 5, 2026, 04:28 AM