dayz-model-pipeline

Pass

Audited by Gen Agent Trust Hub on Oct 26, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and audit external 3D model files, referred to as 'Artist returns' in the 'Artist-return .p3d preflight' section of SKILL.md. This constitutes an ingestion point for untrusted binary data that could theoretically contain malicious instructions in metadata fields.
  • Ingestion points: External .p3d files are read using the py3d library as described in SKILL.md and references/dayz-1-30-model-pipeline.md.
  • Boundary markers: The skill emphasizes 'Preflight' checks (LOD counts, bounding box verification) which serve as structural validation, but it lacks explicit LLM-level boundary markers for content within the binary format.
  • Capability inventory: The skill has significant capabilities including file system access (reading/writing .p3d, .png, .rvmat, .cpp, .cfg files) and execution of external tools (Blender).
  • Sanitization: The pipeline performs extensive geometric and structural sanitization (convexity, mass assigned, axis alignment) to ensure game compatibility, but does not specify textual sanitization for descriptive metadata strings within the processed files.
  • [DYNAMIC_EXECUTION]: The skill generates and executes Python scripts within a headless Blender environment to produce 3D geometry.
  • Evidence: SKILL.md and references/blender-headless.md detail the use of blender --background --python script.py to automate modeling tasks such as primitive creation, modifier application, and OBJ export.
  • [COMMAND_EXECUTION]: The skill instructions include shell commands for setting up the environment and verifying tool versions.
  • Evidence: SKILL.md provides commands for installing system packages (apt-get install -y blender) and Python dependencies (pip install opensimplex, pip install -e tools/py3d), as well as version assertion checks using python3 -c.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 26, 2026, 08:17 AM