Design Memory has 7 tools with basic descriptions but critical gaps in schema documentation and parameter clarity. Tool names follow verb_noun convention (get-*, audit-*, suggest-*, record-*), which is good. However, no visible input/output schemas in the provided source code. Descriptions are present but generic (10-50 chars for most), lacking WHEN to use guidance and dependency hints. Parameters are not documented with types or constraints. Error handling and recovery guidance are absent. The server lacks tool annotations (readOnlyHint/destructiveHint) despite having clear risk classifications (READ_ONLY vs WRITE). No evidence of pagination, result limits, or structured output documentation.

• Add explicit JSON Schema input/output definitions to all 7 tools. Include parameter types (string, number, boolean, enum), descriptions, and constraints. Document return structures with field types and descriptions.
• Expand tool descriptions to 50-150 chars. Add WHEN to use guidance: 'Call get-context first to understand the current design state before auditing changes.' Include prerequisites and dependencies.
• Implement tool annotations: add readOnlyHint=true to all get-* and audit-diff tools; add destructiveHint=true to record-decision. This enables agents to reason about safety and retry logic.
• Add error handling guidance to each tool. E.g., 'If no reference snapshot exists, call sync-reference first.' Return structured error responses with actionable next steps, not raw stack traces.
• For list-returning tools (get-decisions, get-tokens, get-contract), add limit and offset parameters with defaults (limit=20, offset=0). Document result structure including total_count and next_offset for pagination.