Four tools with complete schemas and descriptions, but significant gaps in parameter documentation and output schema clarity. Tool names follow verb_noun convention (list_, get_, edit_, merge_). Descriptions are present (avg 150 chars) but lack actionable context for LLM selection. Critical issue: edit_video's 'editing_details' parameter is an untyped object with inline documentation instead of a proper schema, LLMs cannot validate this. No output schemas documented. Error handling absent. Composition is reasonable (each tool has one responsibility), but parameter descriptions are sparse or missing type constraints.
Edit a video by adding clips or trimming sections. Supports trimming start/end, extracting clips, and updating metadata like title and description.
Retrieve a specific video by ID. Returns full video details including title, description, duration, embed URL, share link, and transcript availability.
Retrieve a list of recorded videos from Loom. Returns video metadata including IDs, titles, thumbnails, and creation dates. Supports pagination and folder filtering.
Combine multiple videos into one merged video. Provide at least 2 video IDs in the desired merge order. Optionally specify a title for the merged result.
edit_video's 'editing_details' parameter is an untyped object with inline documentation. LLMs cannot validate nested fields (trim_start, trim_end, clips, title, description), they lack type info, constraints, and required/optional status. This invites hallucinated payloads.
No output schemas documented for any tool. LLMs cannot plan downstream calls or extract required fields (e.g., video_id from list_recorded_videos to pass to get_video). This breaks tool chaining.
Parameter descriptions lack type constraints and ranges. 'limit' has min/max but no description of what it limits. 'folder_id' has no description. 'video_ids' array has no description of valid ID format or length constraints.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 59 | <=2025-11-25 | v2 |
No error handling or recovery guidance. If a video_id is invalid, the tool will fail silently. No guidance for LLM on what to do next (retry, search, ask user).
Tool descriptions lack context for LLM selection. 'Retrieve a list of recorded videos' does not explain when to call this vs get_video, or what pagination behavior to expect. No mention of result limits or default ordering.