gdoc-claude has 5 tools with reasonable naming (verb_noun pattern) and descriptions present for all tools. However, critical gaps undermine quality: (1) Parameter descriptions are sparse or missing for several inputs (e.g., 'tab' param in gdoc_comment lacks detail on when/why to use it). (2) Output schemas are not documented, callers cannot see what fields to expect from gdoc_read, gdoc_list_comments, or gdoc_comment returns. (3) No error handling guidance, tools fail silently or with generic exceptions rather than actionable recovery steps. (4) No input validation rules stated (e.g., what constitutes a valid doc_url? ID vs full URL?). (5) The 'occurrence' parameter defaults to 1 but lacks bounds or validation description. These gaps force LLMs to guess at tool behavior and error recovery.

• Document output schemas for gdoc_read (return type: string plaintext?), gdoc_list_comments (return type: list of comment objects with fields?), and gdoc_comment (return type: dict with 'anchored' boolean and comment_id?). Include field names, types, and descriptions.
• Add parameter descriptions for 'tab' (e.g., 'Optional tab title to switch to before commenting; if omitted, uses current tab') and 'occurrence' (e.g., 'Which occurrence of target text to anchor to (1-based); defaults to 1 if target appears multiple times').
• Add input validation rules to descriptions: 'doc_url must be a Google Docs URL (https://docs.google.com/document/d/...) or a document ID (alphanumeric string).' Specify character limits and allowed characters for 'target' and 'comment'.
• Add error handling guidance: 'If target text not found, try gdoc_read() to verify the exact text in the doc. If browser is not signed in, call gdoc_login() first. If browser closes, gdoc_status() will return false and the next call will auto-restart.'
• Declare idempotency: 'gdoc_comment is idempotent, calling it twice with the same parameters will create only one comment (or update the existing one if already present).' If not idempotent, explain the duplicate risk and offer a check-before-comment pattern.