Rolester
Guides

Discovering Companies

Grow your target-company set when the sweep returns only roles at familiar names.

The search-jobs sweep runs against a configured set of boards and companies. When that set is exhausted, the same names keep appearing. discover-companies and research-boards are the two skills that grow the inputs.

discover-companies

When you want to find new employers to track:

find companies that are hiring [role] / add companies to my search what companies should I be targeting? the sweep keeps showing the same names

discover-companies web-searches for employers likely to be hiring your target roles, resolves each result to a scannable ATS careers board, legitimacy-screens them (filters staffing farms and no-board results), and proposes adding them to config/sourced-scan.json tracked_companies[].

It is confirm-first: you review the proposed additions and approve before anything is written.

This is the company analog of research-boardsresearch-boards finds boards and aggregators, discover-companies finds employers and wires their specific board into the sweep.

research-boards

When you want new job boards or aggregators:

find more job boards / what sources should I add? / the queue has gone dry

research-boards web-searches for boards relevant to your domain and role families, legitimacy-screens each result, and proposes adding them to config/search-sources.yml.

Again, confirm-first — nothing is written until you approve.

setup-searches

After adding new companies or boards, the sweep picks them up automatically on the next run. To rebuild or tune your search configuration from scratch:

set up my searches / rebuild my search config

setup-searches generates a baseline search config from your role buckets, lets you curate the enabled sources, and imports individual searches (including pasted board URLs with embedded filters preserved).

How the sweep works

Once your search config is in place:

run a search / source new roles / refresh the queue

search-jobs scans your configured sources, dedupes against the existing sourced[] queue and applications[] history, checks link liveness, applies a coarse triage fit score to each new entry (fitScore / fitBucket / fitBasis), and saves the JD bodies locally. Top-scoring sourced roles can optionally be handed to evaluate-job for an immediate body-read gate.

On this page