Single tool 'execute_skill_action' has a minimal schema with only one string parameter. Description is generic (67 chars) and lacks specificity about what 'deterministic, zero-dependency Influencer Marketing skill operations' actually means. No parameter descriptions beyond the bare name. No output schema documented. No error handling guidance. The tool accepts a free-form 'query_payload' string with no validation, constraints, or format specification. No indication of what valid payloads look like, what the tool returns, or how to recover from failures. This is a wrapper tool that delegates to an opaque backend without exposing the actual operations available.

• Replace free-form 'query_payload' with explicit, typed parameters. If the tool schedules campaigns, accept campaign_name, creator_commitments (dict), embargo_windows (bool), etc. Use enums for known values (e.g. campaign_type: ['product_launch', 'seasonal', 'influencer_collab']).
• Document the output schema explicitly. Specify all fields returned (e.g. calendar_schedule_id, campaign_name, total_scheduled_posts_count, embargo_blackout_windows_enforced, optimal_posting_time_utc, interactive_content_calendar_url) with types and descriptions.
• Expand tool description to 100-150 chars explaining: 'Schedule influencer content deliverables and embargo windows for a marketing campaign. Returns a calendar with posting times and interactive URL.' Add parameter descriptions for each input.
• Implement input validation and return structured error responses. E.g. 'Invalid campaign_name: must be 1-100 alphanumeric chars. Got: ""' or 'Missing required creator_commitments dict with format {"@handle": "YYYY-MM-DD"}.'
• Consider splitting into multiple tools if the backend supports distinct operations: schedule_campaign_calendar, update_embargo_windows, get_calendar_url. Each tool should have a clear, specific name and schema.