The Agent Skills Directory

[SAFE]: The skill's core functionality is restricted to local Blender mesh operations. The instructions explicitly prohibit unauthorized uploads, publications, or installations in the provided acceptance checklist, showing a strong focus on user authorization.
[INDIRECT_PROMPT_INJECTION]: While the skill processes external data (source meshes), which represents a theoretical ingestion point, it mitigates risk through several layers: (1) Ingestion Point: Ingests source mesh objects; (2) Boundary Markers: Uses a structured request shape and explicit handover flags for human review; (3) Capability Inventory: Restricts actions to registered retopo mesh commands; (4) Sanitization: Implements retopo.validate to check for geometric anomalies and deviations before results are accepted.

blender-retopology