rip-vehicle-import
Pass
Audited by Gen Agent Trust Hub on Oct 28, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill's primary function involves processing external vehicle assets (ripped models) and associated metadata, which constitutes an ingestion surface for untrusted data. \n
- Ingestion points: The workflow processes external vehicle model data, JSON profiles, and asset contracts (e.g.,
asset-contract.json). \n - Boundary markers: Structural data boundaries are enforced using JSON schemas (
asset-contract.schema.json) to validate the format of incoming decisions and inventories. \n - Capability inventory: The agent is instructed to use powerful automation tools, including headless Blender executions and Python scripts (
rip_p2_import.py,rip_visual_signoff.py), which operate on the ingested data. \n - Sanitization: Integrity is maintained through SHA-256 hashing of artifacts and dedicated checkpoint scripts (
asset_contract_checkpoint.py) that verify conversion lineage before continuing. \n- [COMMAND_EXECUTION]: The skill workflow requires the agent to execute a series of shell commands to automate the conversion and validation process. \n - Evidence: Documentation and cookbooks provide specific command-line strings for running Blender in background mode and executing local Python scripts located within the project's
scripts/directory.
Audit Metadata