Four tools with complete input schemas and descriptions, but significant gaps in parameter documentation and output schema clarity. All tools are READ_ONLY (good), but descriptions lack actionable context for LLM selection. No parameter descriptions beyond the bare names. Output schemas are inferred from stub implementations, not formally documented. Naming is clear (verb_noun pattern), but parameter constraints and error recovery guidance are absent.

• Add descriptions to every parameter. E.g., 'deal_id: The Salesforce Opportunity ID (e.g., 006xx000003DHP). Required.' and 'model: Attribution model to apply. first_touch credits the first touchpoint; u_shaped splits credit between first and last. Choose based on your sales cycle length.'
• Document output schemas in tool descriptions. E.g., 'Returns {deal_id, model, touchpoints: [{channel, date, credit}], total_credit}. Use touchpoints to understand which channels drove the deal.'
• Add error recovery guidance. E.g., 'If deal_id is not found, try search_deals() with the deal name. If the model is unsupported, check the enum values.'
• Expand tool descriptions with WHEN/WHY context. E.g., 'Call this after a deal closes to understand which marketing channels influenced the win. Use the attribution breakdown to optimize future campaigns.'
• Explain enum choices in parameter descriptions. E.g., 'model: Choose first_touch for short sales cycles, u_shaped for complex deals with multiple decision-makers, linear for predictable pipelines.'