The Agent Skills Directory

[PROMPT_INJECTION]: The skill instructions in SKILL.md are focused entirely on technical guidance for Blender's Geometry Nodes API. There are no attempts to override agent behavior, bypass safety filters, or extract system prompts. The instructional tone is professional and context-appropriate.
[DATA_EXPOSURE]: No hardcoded credentials (API keys, tokens, or secrets) were found. While references/procedures.md contains local file paths from the author's development environment (e.g., /Users/emmanuel/...), these are for documentation and testing purposes and do not represent a sensitive data exposure risk for the end user.
[REMOTE_CODE_EXECUTION]: The skill does not download or execute remote code. All operations are performed using the local Blender Python API (bpy). The library in scripts/bx_gn.py provides a declarative builder for node trees, which is a standard and safe use of the Blender API.
[COMMAND_EXECUTION]: The Python script bx_gn.py uses the os module for basic filesystem path manipulation and directory creation to save render snapshots. It does not use subprocess, os.system, or other dangerous command execution functions.
[OBFUSCATION]: The code and instructions are written in plain text. No Base64 encoding, zero-width characters, homoglyphs, or other obfuscation techniques were detected.
[INDIRECT_PROMPT_INJECTION]: As a tool for processing geometry and scene data, the skill has a theoretical surface for indirect injection (Category 8). However, the skill implements rigorous auditing via G.check(md) to detect invalid links or node warnings, and it does not perform any high-risk operations (like network requests or shell execution) based on the geometry data it processes.

blender-geometry-nodes