The Agent Skills Directory

[COMMAND_EXECUTION]: The skill recommends using risky_eval to programmatically verify project metadata, such as the Format.id. This is a standard diagnostic practice within the Blockbench environment to ensure the correct toolset is applied to a project.
[EXTERNAL_DOWNLOADS]: The skill orchestrates the loading of specialized sub-skills like blockbench-modeling and blockbench-texturing. These are vendor-provided modular components used to separate concerns within the 3D modeling suite and do not constitute untrusted external code.
[PROMPT_INJECTION]: The skill processes user intents for 3D modeling and animation, which represents a surface for indirect prompt injection.
Ingestion points: User-provided descriptions for 3D model creation, painting, or animation are ingested from the SKILL.md orchestration instructions.
Boundary markers: The skill defines a strict execution sequence (Classification -> Pre-flight -> Checkpoint -> Execution) which acts as a structural boundary for agent actions.
Capability inventory: The toolset includes content mutation tools (mcp__blockbench__*), file exporting (export_model), and environment inspection tools (risky_eval).
Sanitization: The skill mandates pre-flight state checks (list_outline) and post-execution validation (hytale_validate_model) to verify the integrity of the project before and after changes.

blockbench-use