| name | generate-vacancy-catalog |
| description | Generate or refresh the repository's human-readable vacancy catalog from canonical registry metadata and existing application artifacts. Use after collection, analysis, preparation, or a manual vacancy-status change, and whenever the user asks to list, index, summarize, or rebuild the vacancy catalog. Must run in a separate independent agent or process and may modify only generated files under catalog/. |
Generate Vacancy Catalog
Run the generator only as a separate operating-system process. A normal python run.py catalog invocation satisfies this isolation; do not import and call jobintel.catalog.generate_catalog from the workflow process. No model agent is needed for this deterministic step.
Generate
- Read
AGENTS.md and preserve all user changes.
- Treat
registry/jobs/*/meta.yaml as canonical. The repository uses meta.yaml; never create a parallel metadata.yaml.
- Run
python run.py catalog. This deterministic command scans all vacancies, validates status history, sorts newest first, computes status totals, selects single-file or monthly output, and writes relative links to available artifacts.
- Inspect
catalog/index.md and any generated monthly files. Confirm totals, newest-first ordering, relative links, and missing-artifact markers.
- Confirm the run changed no files outside
catalog/. Report an error rather than repairing or modifying vacancy metadata.
Boundaries
- Never modify
registry/, vacancy content, meta.yaml, job.md, company research, match files, CVs, cover letters, application analysis, or interview preparation.
- Never infer or update a vacancy status. Status changes are manual and belong to the vacancy-management workflow.
- Keep existing status history untouched.
- Do not copy artifact content into the catalog; link to files when they exist and show unavailable artifacts as
—.
- Do not add monthly files unless the deterministic generator selects monthly mode.
