Tachikoma has 17 tools with basic schemas and descriptions, but significant quality gaps prevent production readiness. All tools have descriptions (10-100 chars, mostly 50-80), and input schemas are present with type definitions. However, descriptions are often too terse (e.g., 'List all chat sessions' lacks context on when/why to call), parameter descriptions are minimal or absent, and output schemas are completely undocumented. No error handling guidance, no recovery paths, and no tool annotations. The tool set mixes concerns (session.send handles both text and images; session.respondToApproval conflates approval logic with scope selection). Naming is verb-first but sometimes ambiguous (session.compact, memory.forget). No pagination support despite list tools. Missing critical metadata like required permissions, idempotency guarantees, and side-effect clarity.

• Document output schemas for all tools. Example: session.list should return {sessions: [{id: string, title: string, model: string, createdAt: ISO8601}], total: number}. This enables LLMs to chain calls correctly.
• Expand descriptions to 100-200 chars with context. E.g., 'List all active chat sessions. Call this first to discover available sessions before opening one. Returns session ID, title, model, and creation date.'
• Add descriptions to all parameters. For 'skills' in session.create, document: 'Array of skill names to enable (e.g., ["web_search", "code_execution"]). See engine.listModels for available skills.'
• Add error handling and recovery guidance. For destructive tools, document: 'Deletes the session permanently. No undo. Confirm with the user before calling. Returns {success: boolean, message: string}.'
• Implement pagination: add limit (1-100, default 20) and offset/cursor parameters to session.list and memory.list. Return {items: [...], total: number, hasMore: boolean, nextCursor?: string}.