ShapeLex defines 8 tools with complete input schemas and descriptions. Naming follows verb_noun convention (compress_*, expand, context, inspect, clear, prune). However, descriptions are terse (10-80 chars, well below the 50-200 char LLM-optimized baseline), parameter descriptions lack actionable detail (e.g., 'Session identifier' without explaining when/why to set it), and output schemas are entirely undocumented. Error handling and recovery guidance are absent. The tool set is coherent but descriptions need expansion to guide LLM selection and parameter usage.

• Expand all tool descriptions to 80-150 chars. Example: 'shapelex_compress_messages' → 'Compress older or repeated chat messages into compact sx:// handles to reduce token usage. Use when conversation history grows repetitive or when switching topics. Returns handle and token savings.'
• Document output schemas for all tools. Specify return type (object/array), required fields, and field types. Example: shapelex_expand returns {handle: string, text: string, tokens: number, confidence: string}.
• Add parameter constraints and examples to descriptions. Example: 'budgetTokens (integer, 1-10000): Target token count for compression. Smaller budgets create more aggressive summaries; larger budgets preserve detail.'
• Add confirmation step to destructive tools. shapelex_clear and shapelex_prune should require dryRun: true first, then explicit confirmation before executing. Document this in descriptions.
• Describe shapelex_inspect action enum values. Example: 'action (enum): search=find memory by query, retrieve=get full text of handle, explain=summarize memory, risk=assess if expansion needed, stats=show memory usage.'