OpenImageIO and OpenColorIO skills for DCC texture pipelines
Four tools with complete input schemas and basic descriptions, but significant gaps in LLM-optimization and error guidance. All tools have verb-noun naming (convert_, validate_, inspect_, make_) and documented parameters with types. However, descriptions are minimal (10-60 chars), lack context for tool selection, and omit prerequisites or when-to-use guidance. No output schemas documented. Error handling returns generic exceptions without recovery guidance. Parameter descriptions are sparse, e.g., 'Source color space' lacks format/constraint details. No enums for constrained inputs (color spaces, boolean flags). Tools lack idempotency hints and destructive operation warnings. Composition is reasonable (single responsibility per tool), but chaining requires manual ID extraction.
Convert image color space using OpenImageIO and OpenColorIO
Inspect image metadata and properties using OpenImageIO
Create optimized renderer texture from source image with optional color space conversion
Validate OpenColorIO configuration file
Descriptions are too brief (10-60 chars) and lack LLM-optimization. Missing WHEN to use, prerequisites, and side effects. E.g., 'Convert image color space using OpenImageIO and OpenColorIO' does not explain when to call this vs make_renderer_texture, or that it modifies files.
No output schemas documented. Callers cannot predict response structure (e.g., does 'report' contain raw CLI output or parsed JSON?). LLMs cannot plan downstream tool calls without knowing what fields to extract.
Parameter descriptions lack format/constraint details. 'Source color space' and 'Target color space' do not specify valid values (e.g., 'linear', 'sRGB', 'ACEScg'), ranges, or patterns. LLMs will hallucinate invalid color space names.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 45 | 2026-07-28+ | v2 |
Error handling returns generic exceptions without recovery guidance. E.g., 'Image color conversion failed' does not tell the LLM whether to retry, ask the user, or try a different tool. No actionable error messages.
Destructive operations (convert_image, make_renderer_texture) lack confirmation or dry-run support. No idempotentHint or destructiveHint annotations. Agents cannot safely retry without risking duplicate file overwrites.