Local-first recruitment intelligence and application operations agent with MCP server for typed recruitment and browser tools
RecruitOps Agent exhibits severe definition quality gaps across all dimensions. Tool descriptions are present but generic (avg ~60 chars, well below the 194-char baseline). Input schemas exist but lack parameter descriptions entirely, no tool documents what its parameters control, violating the critical pattern:tool-description rule. 38 tools with zero parameter documentation is a systemic failure. Naming is inconsistent: some tools use clear verb_noun patterns (search_jobs, update_status) but others are vague (browser_observation, crawler_acceptance). No output schemas are documented. Error handling is absent, no recovery guidance, no categorization, no actionable error messages. Security concerns: tools like operation_run and automation_schedule accept generic 'parameters' objects with no validation hints. No tool declares permissions or scope. Composition is poor: multiple tools operate on applications (application_query, application_status_review, application_edit, application_status_update, observe_application_status_page, batch_observe_application_status, verify_application_status_evidence) with overlapping responsibilities and no clear distinction. This server would fail code review by any production standard.
Capture application data from a browser page
Edit application metadata
Query applications by various criteria
Review the current status of a job application
Update the status of a job application
Plan an automation workflow
Schedule an automation workflow
Zero parameter descriptions across all 38 tools. Every parameter (query, company, location, device_id, etc.) lacks documentation of what it controls, expected format, or constraints. LLMs cannot infer parameter semantics from names alone.
No output schemas documented for any tool. LLMs cannot plan downstream calls or extract required fields (e.g., job_id from search_jobs to pass to job_detail). Violates pattern:tool and pattern:tool-chain.
Overlapping application tools with no clear distinction. application_query, application_status_review, application_edit, application_status_update, observe_application_status_page, batch_observe_application_status, and verify_application_status_evidence all operate on applications but lack descriptions explaining when to use each. LLMs will conflate them.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 42 | 2026-07-28+ | v2 |
Disable a scheduled automation
List all scheduled automations
Batch observe application status for multiple applications
Observe and analyze the current state of a browser page
Get the status of a browser operation
Cancel an in-progress browser operation
Describe the capabilities and protocol version of the RecruitOps agent
Get company coverage and recruitment source information
Run a configured crawler for a specific company
Accept or review a crawler run result
Run the daily recruitment synchronization
Get the status of the daily recruitment sync operation
Check the connection status of Edge browser devices
Get detailed information about a specific job posting
Search the knowledge base for relevant information
Observe and analyze an application status page in the browser
Refresh OfferBiu recruitment sources
Run an operational task
Discover public recruitment entries from various sources
Validate a public recruitment entry
Get detailed information about a recruitment email
Process a recruitment email
Get the processing status of recruitment mail operations
Review and analyze a recruitment email
Search recruitment-related emails
Synchronize recruitment emails
Manage schedule operations
Inspect a schedule window for availability
Search for job postings by criteria
Get today's scheduled operations and tasks
Verify evidence for application status changes
Generic parameter names without type suffixes. 'company' could be company_id, company_name, or company_slug. 'status' is ambiguous (HTTP status? application status? boolean?). Violates mxe:natural-identifiers and review:param-validation-rules.
No error handling guidance. Tools like operation_run and automation_schedule accept generic 'parameters' objects with no validation, no error recovery hints, and no actionable error messages. Violates pattern:recovery-guide and pattern:error-classification.