depth-inspect

Pass

Audited by Gen Agent Trust Hub on Oct 6, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructions utilize a local script, scripts/surfaces.py, for processing depth and image data. This script operates on locally supplied parameters and metadata, following a standard and secure tool-use pattern for sensing applications.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes sensor data such as depth maps and photo polygons, which are external data inputs. 1. Ingestion points: Depth maps and polygon definitions referenced in the capture metadata. 2. Boundary markers: Requirement for manual human review of polygons and explicit owner checkpoints for ambiguous data. 3. Capability inventory: Use of the local surfaces.py script for calculating residuals and surface fits. 4. Sanitization: Verification of serialized specification hashes and frame identities to prevent the reuse of outdated or incorrect selection records.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 6, 2026, 11:49 AM