Two tools with basic schemas and descriptions, but significant gaps in parameter documentation, output schema clarity, and error handling. search_jobs lacks actionable guidance on when/how to use it vs import_jobs. import_jobs has incomplete parameter descriptions (e.g., 'Job title' is too generic). Neither tool documents return structure or error recovery paths. No input validation constraints (enums, ranges) visible. Descriptions are adequate length (100-150 chars) but lack LLM-optimized context about dependencies and failure modes.

• Document the return schema for both tools. search_jobs should specify: {jobs: [{title, company, url, location?, description?}], total_count, next_cursor?}. import_jobs should return: {imported_count, failed_count, errors: [{job_index, reason}]}.
• Rewrite search_jobs description to clarify it is a planning tool that guides the LLM to search external job boards, not a direct search API. Add: 'This tool does not search directly, it instructs you to search job boards and then use import_jobs to push results to the dashboard.'
• Add enum constraint to experience_level: z.string().enum(['senior', 'mid', 'junior']).optional() and update description to list valid values.
• Add per-item error reporting to import_jobs. Return {imported: [...], failed: [{index, job_title, reason}]} so the LLM can retry only failed jobs, not all.
• Add error recovery guidance: 'If import fails with "duplicate URL", the job already exists in the dashboard. If it fails with "invalid URL", verify the URL is a valid HTTP(S) link.'

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

• Relies on Stateful initialize / Mcp-Session-Id (removed; protocol is stateless) - make each request self-contained