texturecraft
Pass
Audited by Gen Agent Trust Hub on Nov 17, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data which could potentially contain malicious instructions intended to influence the agent's behavior.
- Ingestion points: The skill ingests complex 3D model files (glTF, FBX, USD, OBJ, Alembic, etc.) and image resources (PNG, JPEG, EXR, HDR) through tools like
mesh_import,project_open, andresource_import(referenced in SKILL.md and references/control-protocol.md). - Capability inventory: The agent is instructed to execute system commands via
texturecraft,texturecraft-cli, andmaterialcraft-clibinaries and has extensive file-writing capabilities, includingproject_save,textures_export, andmesh_export. - Boundary markers: The instructions do not specify the use of delimiters or provide warnings to ignore embedded instructions when processing external data.
- Sanitization: The underlying tool implements workspace-scoped roots (
--read-rootand--write-root) to restrict the agent's filesystem access to specific directories, mitigating risk of path traversal attacks. - [COMMAND_EXECUTION]: The skill is designed to discover and execute local binaries (
texturecraft,texturecraft-cli,materialcraft-cli) to perform its primary functions. It provides detailed logic for searching these executables across different operating systems, including path discovery on POSIX and Windows, and specific search patterns for macOS.
Audit Metadata