Server has 7 tools with complete input schemas and descriptions present. However, descriptions are generic and lack LLM-optimized guidance on WHEN to use each tool or what happens next. Parameter descriptions exist but are minimal (avg ~40 chars vs baseline 72). No output schemas documented. Error handling is basic (throws generic errors without recovery guidance). Tools follow verb_noun naming but lack composition hints for multi-step workflows. No tool annotations (readOnlyHint/destructiveHint) despite clear risk stratification (4 READ_ONLY, 3 WRITE).

• Add output schemas to every tool. Document the structure of listDecks (e.g. {decks: [{name: string, id: number, cardCount: number}]}), listModels, addNote (e.g. {noteId: number, deckName: string}), etc. This enables LLMs to plan downstream calls.
• Expand tool descriptions to 100-200 chars with WHEN/WHY guidance. E.g. 'List all decks in Anki. Call this first to discover available decks before creating notes or searching. Returns deck names and card counts.' instead of 'Get a list of all decks in Anki'.
• Add tool annotations: set destructiveHint=true on createDeck, addNote, addNotes; set readOnlyHint=true on listDecks, listModels, getModel, searchNotes. This signals to agents which operations have side effects.
• Specify the 'options' parameter fully. For createDeck, document valid keys (e.g. 'config' object with 'newSpread', 'resched', 'dyn' fields). For addNote, document 'allowDuplicate' (boolean), 'duplicateScope' (enum: 'deck'|'collection'), and 'duplicateScopeOptions' structure.
• Enhance error handling with recovery guidance. Catch connection errors and return 'Cannot connect to Anki. Ensure Anki is running and AnkiConnect plugin is installed.' For invalid deck/model names, return 'Deck "X" not found. Available decks: [list]. Did you mean one of these?' This guides agent retry logic.