Single tool 'n2_arachne' with comprehensive schema but significant issues. Tool name lacks action verb (should be 'assemble_context' or 'search_codebase'). Description is adequate (127 chars) but generic. Schema is well-structured with Zod types and descriptions for all 14 parameters, but the unified action-dispatch pattern creates ambiguity: LLMs cannot infer which parameters apply to which action without reading implementation. No output schema documented. Error handling returns text responses but lacks recovery guidance or categorization. Parameters like 'action' enum are well-defined, but optional parameters create silent failures (e.g., 'query' required for search but marked optional in schema).

• Split n2_arachne into separate tools: 'assemble_context' (core), 'search_code', 'index_codebase', 'backup_index', 'restore_index', 'get_index_status'. Each tool has one clear action, explicit required/optional params, and documented output schema.
• Document output schema for each action. Example: search returns {results: [{file_path, start_line, end_line, chunk_type, name, score, token_count, content, language}], total_count, backup_results?}. assemble returns {context: string, token_count, layers_used, active_file_context?}.
• Add action-specific parameter validation in descriptions. Example: 'query (required for search/assemble actions, ignored for index/backup/restore/gc)' or split into separate tools to eliminate ambiguity.
• Enhance error messages with recovery guidance. Instead of 'Error: query is required for search action', return 'Missing query parameter. Call search_code with query="<search term>" and optional topK (default 10), language filter, or searchBackups=true.'
• Add per-action examples in tool descriptions. E.g., 'assemble_context: auto-assembles optimal AI context from indexed codebase. Returns markdown-formatted context with token budget tracking. Example: {"action": "assemble", "budget": 40000, "activeFile": "src/main.ts"}'