The Agent Skills Directory

[DYNAMIC_EXECUTION]: The script scripts/zb_stylized.py uses importlib.import_module to dynamically load sibling modules (zb_audit, zb_review, zb_launch) from a relative vendor path. It also contains a build_call function that generates Python code as strings to be executed within the ZBrush environment via a bridge interface.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external sources, specifically OBJ mesh files and PNG canvas renders. The Surface.from_obj and silhouette_mask functions in scripts/zb_stylized.py process these files to derive geometric properties. This represents an attack surface where maliciously crafted geometry or image metadata could potentially influence the agent's behavior.
Ingestion points: scripts/zb_stylized.py (OBJ and PNG files)
Boundary markers: Absent
Capability inventory: File writes, ZBrush command execution, and local network communication
Sanitization: Absent
[COMMAND_EXECUTION]: The skill uses a socket-based bridge (port 7788) to send and execute sculpting commands, UI interactions, and mesh operations inside the ZBrush application process.
[EXTERNAL_DOWNLOADS]: The SKILL.md file contains instructions for the agent to ask the user to add sibling skills using npx from the vendor's official repository scenario-labs/skills.

scenario-zbrush-stylized