mesh-texture-paint
Pass
Audited by Gen Agent Trust Hub on Nov 20, 2026
Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes subprocess calls in
ph3_paint.pyandqwen_paint.pyto interact with a local ComfyUI client (comfy_asset_client.py). The command lines are strictly constructed using absolute paths derived from the script's location, ensuring they point to intended internal components without risk of arbitrary command injection. - [DYNAMIC_EXECUTION]: In
concept_match.py, the skill usesimportlib.utilto load a logic module (embed.py) from a neighboring skill folder (codex-visual). This facilitates functional dependency across skills within the same environment and uses a deterministic relative path based on the script's filesystem origin. - [INDIRECT_PROMPT_INJECTION]: The skill acts as an ingestion point for 3D meshes (GLB) and concept art images (PNG) which are used to guide the AI texturing process. While these represent external data surfaces, the skill limits the agent's capabilities to geometric math, local image rendering, and coordination with a configured ComfyUI server, preventing malicious payload escalation.
Audit Metadata