Single tool 'convert_html_to_asset' has a comprehensive schema with 14 parameters, all properly typed with min/max constraints and enum values. Description is detailed (340+ chars) and explains use cases. However, output schema is not documented, LLMs cannot predict return structure. Parameter descriptions are present but some lack actionable format guidance (e.g., 'output' param says 'must be inside OUTPUT_DIR' but doesn't explain the directory structure or how to discover it). Error handling is not visible in the tool definition. Tool name is clear and verb-prefixed ('convert_'). No security concerns visible (no secrets in params). Composition is single-purpose, appropriate for the domain.
Convert HTML, a URL, a local HTML file, or a local image to an image or animated asset (PNG, JPEG, WebP, GIF, MP4, WebM). Exactly one input must be provided: html, url, file, or image. Examples: - To capture only a specific element: set selector to "#card" - For animated output: set format to "gif" and either use auto: true or set duration - For platform-sized output: set profile to "linkedin-post" instead of width/height Available profiles: linkedin-background | linkedin-post | linkedin-article-cover | linkedin-profile | linkedin-single-image-ad | linkedin-career-background | twitter-post | twitter-header | twitter-ad | twitter-video | twitter-ad-landscape | instagram-post-3-4 | instagram-post-4-5 | instagram-post-square | instagram-story | instagram-reel | instagram-profile | instagram-story-video | square | instagram | twitter | linkedin
Output schema not documented. LLMs cannot predict return structure (file path, dimensions, metadata, error details). Agents cannot plan downstream operations or extract required fields.
Parameter 'output' description lacks actionable guidance. States 'must be inside OUTPUT_DIR' but does not explain how OUTPUT_DIR is configured, what the default is, or how to discover valid paths. LLMs will guess and fail.
No error handling guidance in tool definition. If rendering fails (invalid HTML, timeout, unsupported format), the tool definition does not explain what error messages to expect or how to recover. Agents cannot self-correct.
Mutually exclusive parameters (html, url, file, image) are documented in the description text but not formally declared in the schema. LLMs may pass multiple inputs, causing ambiguous behavior.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 57 | 2026-07-28+ | v2 |
Parameter 'selector' description lacks format guidance. Does not specify CSS selector syntax, valid characters, or examples beyond '#card'. LLMs may pass invalid selectors.