Four tools with mixed quality. All have descriptions and basic schemas, but several critical gaps: (1) capture_note has Korean-language parameter descriptions ('결정 캡처') that are opaque to English-speaking LLMs and violate the chat-data-model principle; (2) parameter descriptions lack actionable constraints (no enums, ranges, or format specs); (3) no output schemas documented; (4) error handling absent, no guidance on what to do if a note fails to save or search returns nothing; (5) brief tool description is vague ('세션 시작 브리핑') and doesn't explain when/why to call it vs search_notes. Naming is reasonable (verb_noun pattern), but parameter design is weak.

• Translate all parameter descriptions to English. Replace Korean descriptions with clear, actionable English text. E.g., 'choice' → 'Decision made (required if capturing a decision; pair with why parameter)'.
• Document output schemas for all tools. E.g., search_notes returns {snippets: [{text: string, label: string, path: string}], total: number}. Include this in tool descriptions or a separate schema block.
• Add enums and constraints to parameters. E.g., volatility: enum ['high', 'low']; limit: {type: 'number', minimum: 1, maximum: 20}; folder: description should list available labels or reference whoami output.
• Clarify brief vs search_notes distinction. E.g., 'brief: Use at session start with a project hint to load past decisions and assumptions. search_notes: Use mid-session to find specific notes by keyword.'
• Add error handling guidance. E.g., 'If folder label is invalid, call whoami to see available labels. If search returns no results, try a broader query or different folder.'