matlab-point-cloud-file-io

Pass

Audited by Gen Agent Trust Hub on Jun 22, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [SAFE]: No security concerns or malicious patterns were identified. The skill provides correct and well-documented guidance for lidar data processing using official MATLAB APIs from a well-known vendor.
  • [PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection as it allows the agent to ingest external point cloud data and subsequently perform file writing operations. This risk is inherent to the skill's primary intended purpose and is handled through standard platform guardrails.
  • Ingestion points: Data reading functions including pcread, lasFileReader, velodyneFileReader, ousterFileReader, hesaiFileReader, e57FileReader, and ibeoLidarReader located in SKILL.md and reference documents.
  • Boundary markers: Absent (standard for binary formats like PLY, PCD, LAS, and PCAP).
  • Capability inventory: File system write access through pcwrite and lasFileWriter + writePointCloud operations.
  • Sanitization: Absent; the skill relies on the Lidar Toolbox's internal format validation and object-based data handling.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 22, 2026, 11:14 AM
Security Audit — agent-trust-hub — matlab-point-cloud-file-io