This server has minimal definition quality. Both tools have descriptions and basic input schemas, but they are extremely brief, lack parameter descriptions, and have no output schema documentation. The tool interface is functional but falls well short of production standards. Naming is reasonable (verb_noun pattern), but parameter documentation is absent entirely. No error handling guidance, no validation messaging, and no structured output definitions.

• Expand tool descriptions to 100-150 characters each. Example for add_note: 'Add a new sticky note to persistent storage. Each note is appended as a single line. Returns confirmation. Non-idempotent: calling twice adds the note twice.'
• Document input parameter constraints: 'note (required, string, max 1000 chars): The text of the sticky note to add. Plain text only.'
• Document output schemas. Example for add_note: 'Returns {"status": "success", "message": "Note added."}' (or return a structured object from the function).
• Document output schema for read_notes: 'Returns {"notes": [list of note strings], "count": integer, "truncated": boolean}' and add optional limit/offset parameters.
• Add error handling with actionable messages. Example: 'Note too long (got {len(note)} chars, max 1000). Truncate and retry.' instead of silently appending.
• For read_notes: implement pagination. Add optional parameters 'limit' (default 20, max 100) and 'offset' (default 0). Return total count so LLM knows if more notes exist.
• Declare idempotency in add_note description. Either state 'Non-idempotent: duplicate calls add duplicate notes' or add a deduplication mechanism.
• Add tool annotations if the MCP framework supports them: add_note is destructiveHint=false (appends, not destructive but mutating); read_notes is readOnlyHint=true.