ArchLens MCP exposes 3 read-only tools with reasonable naming (verb_noun pattern) and descriptions (120-180 chars, within baseline 194 avg). However, input schemas lack critical detail: parameter descriptions are present but sparse, no enum constraints are enforced for 'detail_level' and 'direction' (only documented in description text, not schema), and output schemas are completely undocumented. The 'process_name' parameter is optional but lacks guidance on discovery. Error handling and recovery paths are absent. Composition is sound (each tool has one responsibility), but schema completeness and LLM-actionable error guidance are missing.

• Add explicit JSON Schema enum constraints for 'detail_level' and 'direction' parameters instead of relying on description text alone.
• Document output schemas for all 3 tools. Example: archlens_architecture returns {modules: [{name, dependencies, stats}], layers: [...], techStack: [...]}. This enables LLMs to chain results into follow-up calls.
• Add error recovery guidance: 'If process_name is not found, call archlens_process with no arguments to list available processes.'
• Specify result limits in descriptions: 'Returns up to 50 modules. For full list, use detail_level=full.' This prevents context window exhaustion.
• Add parameter descriptions for 'show_algorithms' explaining what algorithm details include and when they are useful (e.g., 'Set to false to reduce output size for large processes').