Server has 9 tools with basic descriptions and input schemas. Naming follows verb_noun convention (connect, health, list_folders, latest_emails, read_email, search_emails, send_email, mark_read, mark_unread). Most tools have descriptions (10-100 chars), but several are terse and lack context on WHEN to use them or WHAT they return. Input schemas present for most tools with type definitions, but output schemas are not documented in the source. Error handling exists (_friendly_error) but is not surfaced in tool descriptions. No tool annotations (readOnlyHint/destructiveHint) despite clear risk distinctions (READ_ONLY vs WRITE). Parameter descriptions are minimal; e.g., 'limit' lacks min/max bounds. No pagination support despite list_folders and latest_emails returning potentially large result sets.

• Add output schema documentation to all tools. E.g., latest_emails should document: returns array of {message_id: string, subject: string, from: string, date: string, preview: string}. This enables LLMs to chain tools correctly.
• Expand tool descriptions to 100-200 chars. Include WHEN to use (e.g., 'Call this after connect() to verify authentication') and WHAT it returns (e.g., 'Returns list of folder names'). See pattern:tool-description baseline.
• Add tool annotations via FastMCP. Mark send_email, mark_read, mark_unread with destructiveHint=true. Mark read-only tools with readOnlyHint=true. This enables agents to reason about retry safety and side effects.
• Document parameter constraints in descriptions. E.g., 'limit: Maximum number of results (1-100, default 10)' and 'query: IMAP search syntax (e.g., FROM [email protected], SUBJECT keyword, SINCE 2024-01-01)'.
• Add pagination to list_folders and latest_emails. Accept offset/limit or cursor parameters and return total_count. Cap results at 50 items per call to avoid context window exhaustion.

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

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