Single tool 'execute_skill_action' has a vague, marketing-focused description (97 chars) that fails to explain WHAT the tool does, WHEN to use it, or what it returns. Input schema exists but is minimal: single string parameter 'query_payload' with no format constraints, length limits, or examples of valid payloads. No output schema documented, response is a generic JSON echo with no structured fields the LLM can rely on. Tool name uses 'execute' (generic verb) + 'skill_action' (vague noun), violating verb_noun clarity. No error handling guidance. No parameter validation rules. Composition is poor: tool claims to handle 'Autonomous CRM, B2B Sales Intelligence & Lead Enrichment', three distinct domains in one tool, violating single-responsibility principle.

• Rename tool to a specific action verb + noun pair. Examples: 'enrich_lead', 'search_prospects', 'score_opportunity'. Avoid generic verbs like 'execute' or 'process'.
• Rewrite description to 50 - 150 chars explaining: (1) What the tool does, (2) When to use it vs. other tools, (3) What it returns. Example: 'Enrich a prospect record with company data, funding events, and decision-maker contact info. Returns structured prospect object with enrichment metadata.'
• Add format constraints to 'query_payload': specify expected JSON structure, required fields, and length limits. Or split into separate parameters (prospect_email, company_name, enrichment_type) with enums and descriptions.
• Document the output schema with typed fields: prospect_id (string), company_name (string), enrichment_score (number 0 - 100), detected_triggers (array of strings), etc. Include which fields are always present vs. optional.
• Split the tool into three focused tools: 'enrich_lead' (prospect data), 'search_company_intelligence' (company catalysts), 'score_sales_opportunity' (lead scoring). Each should have a clear, single responsibility.