Local-first desktop workspace for agent-assisted job search and application drafting
Applyer exhibits significant definition quality gaps. Tool descriptions are present but generic (avg ~50 chars, well below the 194-char baseline). Input schemas are partially visible but lack rigor: most parameters have minimal type info and descriptions. Critical tools like fill_application and edit_application are inferred from UI hints rather than explicit MCP server registration, capping their scores at 50. Error handling is absent, no recovery guidance, no categorization. No tool annotations (readOnlyHint/destructiveHint) despite clear WRITE vs READ_ONLY risk patterns. Output schemas are undocumented. The server lacks composition guidance (e.g., which tools chain together) and parameter constraints (enums, ranges). Overall, this reads as a prototype-stage tool set, not production-ready.
Edit an existing job application
Exclude a job from future consideration
Fill out and submit a job application form
Fetch detailed information about a job posting
Inspect the form fields of a job application
List all tracked jobs in the local database
Queue a job for tracking and application
Search for job postings across multiple job boards
fill_application and edit_application are inferred from UI hints (toolPayloadHints.ts), not explicitly registered in MCP server code. Tool definitions are not directly visible in source.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite clear risk classification. WRITE tools (queue_job, fill_application, edit_application, exclude_job) lack destructiveHint; READ_ONLY tools lack readOnlyHint.
Descriptions are generic and under 50 chars (e.g., 'Queue a job for tracking and application'). No guidance on WHEN to use each tool, prerequisites, or what the LLM should expect. Baseline is 194 chars.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 40 | 2026-07-28+ | v2 |
Output schemas are not documented. LLMs cannot plan downstream calls or extract required fields (e.g., what does search_jobs return? job IDs? URLs? company names?).
No error handling guidance. Tools lack recovery hints (e.g., 'If job not found, try search_jobs first'). No error categorization (retryable vs user-fixable vs fatal).