The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external texture filenames and mesh attributes, presenting a surface for potential instruction injection if an agent interprets these strings as commands. Ingestion points: bx_materials.py extracts roles from texture filenames in find_texture_set() and reads vertex attribute names from the Blender mesh object in read_vertex_values(). Boundary markers: There are no specific delimiters or safety instructions provided to the agent to differentiate these data values from operational instructions. Capability inventory: The skill allows writing files to disk via save_image() and direct manipulation of the Blender environment through the bpy API. Sanitization: Extracted identifiers are not sanitized before processing.
[DYNAMIC_EXECUTION]: The skill uses runtime path modification to resolve local dependencies. Evidence: In bx_materials.py, the script calculates the location of a sibling blender-expert repository and appends it to sys.path. This enables the import of internal modules bx_gui and bx_review during execution.
[DATA_EXFILTRATION]: The skill's documentation and source code contain hardcoded absolute paths that reveal information about the author's local file system. Evidence: Absolute paths, such as /Users/emmanuel/Developer/pro/2026-09-24 Blender Expert Skills/, appear in SKILL.md and bx_materials.py, identifying the developer's username and project organization.

blender-texturing-shading