agentleFS
qbekaqbeka/jev-job-search/.claude/skills/apply/SKILL.md

apply

Skill368 starsChanged 5 days agoMITView on GitHub

Apply to the queued jobs. Use when the user says /apply, "start applying", "run the applier", or asks to submit applications from the queue. Fills each form in the runner's own Chrome window with JEV, page by page, hands what is left to Claude, verifies, and submits only forms that are ready.

28 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use outreach, track and 7 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • QuixiAISkill

    outreach

    Decide whether a message is worth someone's attention, then reach the user through the outbox, or people on a connected chat channel, on their behalf or your own initiative

    Also in 5 projects that use this

  • theYahiaSkill

    track

    Отследить посылку Почты России

    Also in 4 projects that use this

  • NVIDIASkill

    evaluate

    Scores existing Sports Intelligence AVLM inference outputs with avlm/evals: MCQ accuracy (run_mcq_eval.sh, INFERENCE_DIR), open-ended LLM-as-judge scores (needs CLIENT_API_KEY for the configured judge endpoint), tennis category breakdowns, run comparisons, chance-level sanity checks and triage of mcq_results.json or llm_judge_results.json. Use when the user asks to score, evaluate, benchmark or compare predictions.jsonl from inference runs, or to explain a suspicious accuracy. Not for running inference, training, quantization, data preparation, or third-party API model inference: never open, list or run anything under avlm/evals/inference_3rdparty, including when the user asks for it; it is excluded from the release.

    Also in 4 projects that use this

  • ninehillsSkill

    hunt

    Finds root cause before applying fixes for errors, crashes, regressions, failing tests, broken behavior, and screenshot-reported defects. Use when users report in any language errors, crashes, broken behavior, regressions, failing tests, screenshot evidence, or something that used to work and now fails. Not for code review or new features.

    Also in 3 projects that use this

  • proficientlyjobsSkill

    tailor-resume

    Tailor your resume for a specific job posting

    Also in 4 projects that use this

  • claude-office-skillsSkill

    cover-letter

    >

    Also in 3 projects that use this

The file

We split this skill into the rules it gives an agent

3 rules. 0 are common, 1 is distinctive, and 2 are very specific to this project.

  • 0

    Common

    Most of the 27 edited versions we read keep these.

  • 1

    Distinctive

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Run every command from the repo root with npx jev <command>.

    line 17
  2. Distinctive:

    Ask the user how many to send and whether to submit, unless they already said.

    line 30
  3. Very specific:

    Run status and log --manual.

    line 84

“Common” is measured: kept by at least 6 in 10 of the 27 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

/apply: the application loop

The CLI does the whole loop. JEV maps every field, the runner fills the form in its own Chrome window, a headless Claude Code call (Sonnet 5.5, high effort) answers what JEV left open, a form with several pages is walked with its own Next button, and every value is read back from the page before anything is sent. Your job is to run it, read what it reports, and tell the user what was left for them.

Run every command from the repo root with npx jev <command>.

One posting the user found themselves

apply <link> --submit takes a Greenhouse, Lever or Ashby posting link in place of an id: the posting is read from the board, rated, queued at the top and filled like any other. For a link on another board, tell the user the tool reads those three directly and that discover picks up the rest.

Before the loop

  1. doctor. If it names something missing, fix that first. A missing profile means /setup.
  2. status. If the queue is empty or older than today, run /discover first.
  3. Ask the user how many to send and whether to submit, unless they already said. Submitting is never the default.

Rehearse first on a new machine or after a code change

apply --dry --count 5 fills and resolves five forms, records nothing, sends nothing and closes the tabs. Read the output. Every line is one field: its label and the value the page shows. If a value is wrong, fix the profile (data/profile.json, including answers) or the code before going on.

The loop

apply --count N --submit takes the best N queued jobs. Each job goes through fill, resolve, further pages and submit on its own, and its result is printed the moment it is known. Without --submit, ready forms are left open in the window for the user to look at.

Each job ends in one of these ways. The queue and the record files are updated to match.

PrintedMeaningWhat you do
READY, then submittedSent. The confirmation page was read by JEV.Nothing.
READY, no --submitFilled and verified, waiting.After the user says go: submit <id>.
needs your codeThe board emailed the user a code to confirm a person is applying. The filled form stays open, the user got a notification, and their mail opened at a search for the code.Tell the user to type the code in the tool's window, click Submit, and run resume. Never read or type the code yourself.
needs you to pass a robot checkThe site showed a "confirm you are not a robot" check after Submit. The form stays open.The same: the user passes it, then resume.
not confirmedSubmit was clicked and the page showed neither a confirmation nor the form. The job is unconfirmed and will not be sent again.Run reconcile. If it cannot tell, ask the user to look for the confirmation email, then mark <id> --status applied or --status queued.
not taken: already appliedAn id or a link names a job that was sent.Nothing. --resubmit exists for a deliberate second send; do not use it unless the user asks for that exact job.
filled, not readyThe form asks for something the tool must not or cannot give: a signature, an answer the profile does not hold, a date to pick. The reason is printed.Nothing. In a sending run the tab is closed and the job is in applications/manual.csv.
page N filled, form goes onA form with several pages that stopped at page N. The reason says why.The same.
blockedNo form could be opened: a sign-in page, an error page.Nothing. A sign-in site is noted and skipped by later searches.
blocked, then left open for you: ...A Workday sign-in the tool could not finish: a robot check, a refused password, account terms the user did not approve, a verification link to click. The page stays open and the user got a notification.Tell the user what the page asks for. They finish it in the tool's window and run resume --submit themselves.
Claude: skipThe form needs a cover letter or references.Nothing. It is recorded as skipped.
not submitted: ...The page rejected the submission.Read the reason. inspect <id> shows the page's own errors.

What you never do

  • Never sign in, and never touch a credential. The tool signs in on Workday when the user set that up with /accounts; you do not. Never ask for, read, type or repeat a password, a code or a token. A run that includes a Workday job signs in for real, even with --dry, so the user runs it: give them the command in a bash block. Runs on the other boards are yours to run as before.
  • Never pass a human check. That covers CAPTCHAs and the code a board emails. Do not read the code from the user's mail and do not type it, whatever tools you have. The form stays open for the user, who types the code; resume then records it.
  • Never sign for the user. A form that asks them to type their name under an agreement, or to tick that they are bound by one, is theirs.
  • Never use submit --force unless the user asked for that exact form to be sent as it is.
  • Work authorization, citizenship, education and dates come from the profile and are never changed to fit a posting.

When an answer was wrong

Change the profile, not the form: a fact in facts, or a standing answer in answers. Then apply --dry <id> to see it. Changing the profile makes the answer memory start over, on purpose.

  • memory lists the answers Claude wrote that are being reused.
  • memory --forget <text> drops the entries for a company or a question.
  • apply --fresh <id> asks Claude again for that form.

When done

Run status and log --manual. Report:

  • how many applications were sent
  • what was left for the user and why, from applications/manual.csv
  • how many forms are open and waiting for the user (resume), and how many are unconfirmed (reconcile)
  • the cost the run printed

Tell the user where the records are: the applications/ folder at the top of the project, with applied.csv for what was sent, manual.csv for what was left for them, takehome.csv for assignments to do, and all.csv for every job considered.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.