discover
Refresh the job queue without a browser. Use when the user says /discover, "find jobs", "refresh the queue", or before an /apply run when the queue is stale. Pulls every source, filters, rates with JEV, writes data/queue.json and applications/all.csv.
Install
npx skills add https://github.com/qbeka/jev-job-search/tree/main/.claude/skills/discover
claude plugin marketplace add https://llmmart.ai/marketplace.json && claude plugin install qbeka-jev-job-search@llmmart
git clone https://github.com/qbeka/jev-job-search.git
The skills CLI installs just this skill, for any of its supported agents. Claude Code installs the whole qbeka/jev-job-search collection as a plugin from our marketplace. Git is the plain clone.
Skill manifest
/discover: build the queue
No browser needed. Run from the repo root:
npx jev discover
It prints counts and the top 50 queued jobs with scores. Takes about a minute. The first search of a day costs about six cents of JEV. A later one costs less, because a posting that has not changed keeps its rating.
Then show the user the top 15 as a short table (score, company, title,
location, posted) and one line of totals. Mention jobs that were skipped for
a reason they might want to override, grouped by reason, using
npx jev status.
Options:
--no-boardsskips polling company boards (faster, fewer fresh postings).--limit Nrates only N new jobs, for a quick check.
To include the user's own tracking sheet, they export it as CSV into
data/imports/ first; every file there is read on the next discover.
If it fails with a TYPESAFE_API_KEY or OPENROUTER_API_KEY error or a missing profile, run
npx jev doctor and follow its next step, or use /setup. If a source fails, the run continues without it and says so
on stderr.
Files (jev-job-search)
-
SKILL.md 1.3 KB
--- name: discover description: Refresh the job queue without a browser. Use when the user says /discover, "find jobs", "refresh the queue", or before an /apply run when the queue is stale. Pulls every source, filters, rates with JEV, writes data/queue.json and applications/all.csv. model: claude-sonnet-5-5 effort: high --- # /discover: build the queue No browser needed. Run from the repo root: ``` npx jev discover ``` It prints counts and the top 50 queued jobs with scores. Takes about a minute. The first search of a day costs about six cents of JEV. A later one costs less, because a posting that has not changed keeps its rating. Then show the user the top 15 as a short table (score, company, title, location, posted) and one line of totals. Mention jobs that were skipped for a reason they might want to override, grouped by reason, using `npx jev status`. Options: - `--no-boards` skips polling company boards (faster, fewer fresh postings). - `--limit N` rates only N new jobs, for a quick check. To include the user's own tracking sheet, they export it as CSV into `data/imports/` first; every file there is read on the next discover. If it fails with a TYPESAFE_API_KEY or OPENROUTER_API_KEY error or a missing profile, run `npx jev doctor` and follow its next step, or use `/setup`. If a source fails, the run continues without it and says so on stderr.
Comments (0)
Sign in to join the conversation.
Reviews (0)
No reviews yet.
No comments yet.