Loading ToolBench
Sync Plaud recordings (AI summary + speaker transcript) into an Obsidian vault as Markdown notes, via the official Plaud MCP.
Three tools with basic schemas and descriptions, but significant gaps in parameter documentation and output schema clarity. All tools are READ_ONLY and well-named with action verbs (list_, get_). Descriptions are present but brief (18-52 chars), below the 50-200 char LLM-optimized baseline. Input schemas exist but lack depth: list_files has minimal parameter descriptions; get_note and get_transcript lack descriptions for some parameters. No output schemas documented. Error handling not visible in tool definitions. Pagination support is present (list_files, get_transcript) but not fully specified.
Fetch AI-generated summary note for a recording
Fetch transcript segments for a recording with cursor-based pagination
List Plaud recordings with pagination support
Tool descriptions are too brief (18-52 chars vs. 50-200 char baseline). 'List Plaud recordings with pagination support' and 'Fetch AI-generated summary note for a recording' lack context on WHEN to use each tool and what data structure is returned.
Output schemas are not documented. LLMs cannot plan downstream calls or extract fields without knowing what list_files, get_note, and get_transcript return. No field names, types, or structure visible.
Parameter descriptions are missing or minimal. get_note's 'file_id' has a description, but get_transcript's 'block' parameter lacks context on valid values (only example 'transaction' given). list_files' 'page_size' has no guidance on valid range.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 56 | 2026-07-28+ | v2 |
No enum constraints for constrained parameters. get_transcript's 'block' parameter should declare valid block types as an enum (e.g. ['transaction', 'speaker', ...]) rather than relying on LLM inference from a single example.
Pagination guidance incomplete. list_files accepts page/page_size but no total_count or next_cursor documented. get_transcript has cursor-based pagination but no guidance on when cursor is exhausted or how many results to expect.