Mailcomb has 6 tools with clear verb-noun naming (search, get_message, list_labels, send_message, add_label, remove_label). All tools have descriptions (avg ~80 chars, within baseline 34-392). Input schemas are present with typed parameters and descriptions. However, output schemas are not documented in the source, parameter descriptions lack detail on constraints/formats, and error handling guidance is absent. No tool annotations (readOnlyHint/destructiveHint) despite clear risk classifications. Composition is sound, tools are single-responsibility and chainable (e.g., search → get_message → add_label). Missing: enum constraints on parameters, pagination details for search results, and recovery guidance in error cases.

• Document output schemas for all tools. Example: search returns [{message_id, subject, from, date, snippet}], get_message returns {message_id, subject, from, to, cc, bcc, body, attachments: [{filename, size, mime_type}]}. This enables chaining and field extraction.
• Add tool annotations via rmcp macros: #[tool(readOnlyHint=true)] for search/get_message/list_labels, #[tool(destructiveHint=true)] for send_message/add_label/remove_label. Agents will respect operation safety.
• Enhance parameter descriptions with constraints. Example: 'Search query string (max 1000 chars; supports field:value syntax like from:[email protected], subject:urgent)'. 'Maximum results (1-100, default 20)'.
• Add error recovery guidance. Example for get_message: 'If message not found, verify the message_id with search() first. Common causes: message was deleted, or ID is from a different account.'
• Implement dry-run for send_message: add optional 'dry_run: bool' parameter. When true, validate recipients and return a preview without sending. Agents can confirm before executing.

Inferred effective spec: <=2025-11-25.

• Relies on Logging (deprecated) - log to stderr or use OpenTelemetry