Loading ToolBench
MCP server for a job search dashboard system that manages job postings, applications, profiles, and search runs with tools for reviewing jobs, preparing applications, and managing job search workflows.
Server has 9 tools with explicit registration, clear naming (verb_noun pattern), and documented schemas. However, descriptions are inconsistent in depth, parameter descriptions are sparse, and output schemas are not formally documented. Most tools lack error handling guidance. Naming is strong (get_, list_, set_, run_, prepare_), but parameter documentation and response structure clarity need improvement. Average tool quality is fair-to-good, but lacks the rigor expected for production agent use.
Generate/reuse tailored resumes and cover letters for top scoring matches. Applications remain manual-submit only.
Return profile, review queue, application, and latest run counts for the Job Search OS.
Get full job details, match explanations, generated materials, and applications for a job.
Return run status, stats, progress messages, and errors for a job search run.
List known jobs with optional status, score, company, location, and profile filters.
List jobs that need review, sorted by match score.
Generate/reuse tailored resume and cover letter, create a ready_to_apply application, and keep manual submission required.
Output schemas not documented. Tools return data structures (profiles, matches, applications, latestRun) but LLMs cannot see what fields to expect. Downstream tool chaining and field extraction require guesswork.
Parameter descriptions missing or minimal. 'runId' in get_search_run lacks detail on format/length. 'appUrl' in prepare_application_package has format='uri' but no guidance on when to override default. LLMs cannot infer intent from sparse descriptions.
No error handling guidance. Tools lack recovery hints. If run_job_search fails, LLM has no guidance on retry strategy, partial completion, or fallback actions. No categorization of retryable vs fatal errors.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 57 | 2026-07-28+ | v2 |
Start or run the scheduled job ingestion pipeline across enabled profiles and sources.
Approve, reject, save, archive, or otherwise update a profile-specific job match status.
Destructive operations lack confirmation. set_job_match_status and prepare_application_package modify state without dry-run or confirmation step. Agents can accidentally reject jobs or generate materials without review.
Pagination limits not enforced in descriptions. list_jobs and list_review_queue accept limit params but descriptions do not state the hard cap (100 and 100 respectively). LLMs may request excessive results.