Lex Intel provides 4 read-only tools for Chinese AI intelligence discovery. All tools have clear, descriptive docstrings and well-structured input schemas. However, there are significant gaps in parameter descriptions, output schema documentation, and error handling guidance. The schema quality varies: lex_search_articles has excellent parameter descriptions and constraints (enum for category, min/max for limit/relevance); lex_get_briefing has minimal parameter descriptions; lex_get_signals lacks output schema documentation; lex_list_sources has no parameters but no explicit output schema. All tools are read-only (safe), but lack actionable error messages and recovery guidance. The codebase shows thoughtful design (e.g., semantic search, time-windowed queries, filtering) but falls short of production-grade tool definition standards expected for agent composition. Parameter descriptions average ~60 chars (below the 72-char baseline for A+ tools); tool descriptions average ~180 chars (within range but some lack usage context). Output schemas are inferred from code but not formally documented in tool metadata.

• Add formal JSON Schema output definitions to all tools. Example for lex_search_articles: {"articles": [{"type": "object", "properties": {"source": {"type": "string"}, "english_title": {"type": "string"}, ...}}], "total": {"type": "integer"}, ...}. This makes output structure machine-parseable and prevents agent hallucination about field names.
• Add min/max constraints to numeric parameters in JSON Schema, not just prose. lex_get_signals should declare: {"days": {"type": "integer", "minimum": 1, "maximum": 30}}. This lets the MCP framework validate before calling the tool.
• Document error cases with recovery paths for each tool. Example: 'If search returns no results, try: (1) removing filters (category, min_relevance), (2) using shorter/broader keywords, or (3) calling lex_get_signals for recent high-relevance articles instead.'
• Expand lex_list_sources description to define the output structure. Example: 'Returns a list of sources with fields: name, url, last_scrape_time (ISO 8601), status (active|error|offline), error_message (if status=error).'
• Add explicit guidance for default/optional parameters. lex_get_briefing: 'Omit the date parameter to fetch today\'s briefing. Provide date=YYYY-MM-DD to retrieve a specific day. Returns an empty briefing if none exists for that date; try omitting date to get the most recent.'
• For lex_search_articles, document pagination behavior and result limits. Currently returns 'has_more: bool' but no next_cursor or offset mechanism. If results are truncated, provide a way for the agent to fetch the next page.

Inferred effective spec: <=2025-11-25.

• Relies on Logging (deprecated) - log to stderr or use OpenTelemetry