spine-rig-animation
Pass
Audited by Gen Agent Trust Hub on Nov 15, 2026
Risk Level: SAFEDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [DYNAMIC_EXECUTION]: The skill documentation in references/pipeline-operations.md describes a workflow involving local Python extensions through the ProjectAdapter class. It explicitly notes that these callbacks (local_transform) are executed in a non-sandboxed environment, which is a significant capability that relies on the user providing trusted local scripts.
- [INDIRECT_PROMPT_INJECTION]: The agent is instructed to ingest and process external Spine JSON project files, metadata, and asset records, which constitutes an attack surface for indirect prompt injection.
- Ingestion points: SKILL.md and references/mesh-weights.md describe reading existing Spine JSON/project versions, skeleton schemas, and mesh topology data.
- Boundary markers: The instructions utilize tool schemas and explicit validation steps to define the scope of data processing, although no natural language delimiters for untrusted content are specified.
- Capability inventory: The skill utilizes tools for file export (spine_export), project validation (godot_validate), and local Python-based data transformations via the ProjectAdapter.
- Sanitization: The documentation mandates the use of skeleton_validate, mesh_inspect, and job_validate to verify data integrity and compliance with Spine 4.3.x standards before modifications are performed.
Audit Metadata