Server has 4 tools with complete Zod schemas and descriptions. Naming follows verb_noun pattern (write_*, plan_*). Descriptions are present (194-250 chars, within baseline 34-392 range) but lack actionable context for LLM selection. Parameters are typed and mostly described, but lack constraints (enums, ranges, patterns). No output schema documentation. Error handling is absent, LLM receives raw LLM-generated text with no structure, validation, or recovery guidance. Tools are composition-heavy (write_product_hunt_post combines tagline + description + maker comment + checklist) rather than single-responsibility. No idempotency guarantees or retry guidance.

• Add output schema documentation. Define structured response with fields like {sections: [{title, content}], metadata: {word_count, estimated_read_time}}. Return JSON, not raw text.
• Implement error handling with recovery guidance. Catch Groq API errors and return {error: 'API timeout', retry_after: 30, fallback: 'Use template at...'}. Guide LLM on next steps.
• Add parameter constraints: product_name (1-100 chars), product_description (10-500 chars), key_features (enum or comma-separated list with 3-5 items). Validate before calling Groq.
• Split composite tools. Separate write_product_hunt_post into write_ph_tagline, write_ph_description, write_ph_maker_comment, write_ph_launch_checklist. Each does one thing.
• Enhance descriptions with WHEN/WHY. E.g., 'Use this to generate Product Hunt launch copy when you have product details but no PH experience. Returns tagline, description, and engagement strategy. Call this BEFORE write_landing_page_copy to align messaging.'