Three tools with basic descriptions but critical schema gaps. search_jobs and save_job lack visible input schemas entirely; get_job has minimal schema (only 'id' documented). Descriptions are present but generic (10-50 chars for search_jobs, ~80 for get_job/save_job). No parameter descriptions beyond bare names. No output schemas documented. No error handling guidance. Tool names follow verb_noun pattern (positive), but lack the depth required for production LLM agents.

• Add explicit input schemas for search_jobs (e.g., criteria object with fields like skills[], location, remote_only, salary_min) and document all parameters with type, description, and constraints.
• Document output schemas for all three tools. E.g., search_jobs returns [{id, title, company, location, remote, salary, skills}]; get_job returns {id, title, company, location, remote, salary, skills, description}; save_job returns {success, job_id, saved_at}.
• Expand search_jobs description to 50-150 chars explaining WHEN to call it (e.g., 'Search for recently posted jobs matching your stored criteria. Returns summaries with title, company, location, and skill tags. Use before get_job to enrich details.').
• Add parameter descriptions: 'fitReason' should specify format (e.g., 'One-sentence explanation (max 200 chars) of why this job matches the candidate's profile and goals.').
• Add error handling guidance: 'If job not found, try search_jobs to verify the ID exists. If save fails due to invalid fitReason, ensure it is under 200 characters and references the candidate's actual skills.'

Inferred effective spec: 2025-06-18+.

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