jobsearch-mcp has 14 well-named tools with clear verb-noun patterns (search_jobs, track_job, score_fit). All tools have descriptions (avg ~120 chars, within baseline 34-392 range). Input schemas are present and typed for all tools. However, output schemas are NOT documented, the code shows tool implementations but no explicit return type documentation visible in the source. Parameter descriptions are present but often minimal (e.g., 'Job title or keyword to search for' lacks format/constraint details). Error handling is basic, no recovery guidance or actionable error messages visible. No tool annotations (readOnlyHint/destructiveHint) despite clear risk classifications in the spec. Missing pagination guidance for list-returning tools like get_tracked_jobs.

• Add explicit output schemas to all tools. Document return types as JSON Schema (e.g., search_jobs returns {results: [{url, title, company, source}], total_count: number}).
• Implement tool annotations in FastMCP registration: use readOnlyHint=true for read tools, destructiveHint=true for delete_user_profile, idempotentHint=true for idempotent operations.
• Enhance parameter descriptions with format constraints. E.g., 'Location filter (optional, format: city, state or postal code, e.g., "San Francisco, CA")' and 'Job title (1-100 chars, alphanumeric + spaces)'.
• Add error recovery guidance to tool descriptions. E.g., 'If job URL not found in tracking database, call search_jobs to locate it first, then track_job to add it.'
• Add limit and offset/cursor parameters to get_tracked_jobs. Document max results (e.g., 'Returns up to 50 most recent jobs; use offset for pagination').