job-pipeline
Skill0 starsChanged 18 days agoMITView on GitHub
Run a high-volume, no-duplicate job search as a three-worker pipeline (daily Searcher, Analyst, Materials) that shares one tracker sheet, one facts file and one JD folder, with the candidate approving every application. Use when setting up or operating a job search, screening new roles, or deciding what a worker may change.
In this repository4 other files it gives its agents.
- job-fact-checkSkill
- job-form-answersSkill
- job-outreachSkill
- resume-builderSkill
The file
We split this skill into the rules it gives an agent
1 rule. About 0 are common, 0 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈0
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 22
Only the candidate sets Approved, Skipped (except clear eligibility failures) and Applied.
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Job application pipeline
Three linked workers pass roles along one status column. The candidate approves and submits; workers never apply, message, sign in or create accounts.
Searcher (daily) ──New──▶ Analyst ──Screened──▶ Candidate ──Approved──▶ Materials ──Materials ready──▶ Candidate submits ──Applied──▶ …Shared state
- Tracker (Google Sheet or spreadsheet): columns and owners in
templates/tracker_columns.md. Each worker writes only its own columns, so parallel workers never overwrite each other. - Facts file: the only source of truth for the candidate's experience. See
job-fact-check. - JD folder:
jds/YYYY-MM-DD_Company_Role.md, full posting text plus URL, date checked, location rule and pay. - Outputs: one folder per role with
<First>_<Last>_Resume.pdf, optional<First>_<Last>_Cover_Letter.pdf,outreach.md,form_answers.md.
Status flow
New → Screened → Approved / Skipped (candidate) → Materials ready → Applied (candidate) → Screening / Assessment / Interviewing / Offer / Rejected / Withdrawn.
Only the candidate sets Approved, Skipped (except clear eligibility failures) and Applied.
Workers
- Searcher – a scheduled task built from
templates/searcher_prompt.md. Finds recently posted roles that pass the hard filters (location eligibility at role level, pay floor, excluded role types), dedupes against the tracker by company + title and by URL, saves the JD, appends rows asNew. Targets a minimum daily count and broadens keywords if short. - Analyst – for each
Newrow: role type, fit (A = tailor fully, B = base resume with keyword tuning, C = long shot), key gaps against confirmed facts only, suggested base resume, a short JD fingerprint (to detect reposted or changed JDs), thenScreened. Questions that change eligibility or truthfulness go to the candidate, prefixedQUESTION:. - Materials – for each
Approvedrow: build the resume (and CL when required or fit A) withresume-builder, outreach for fit A withjob-outreach, form answers withjob-form-answers; every document passesjob-fact-check; thenMaterials ready.
Coordinator
One session starts the Analyst and Materials sessions, relays batches, and gives the candidate one summary table per batch to approve. Rules live in a PIPELINE.md in the candidate's workspace (start from templates/PIPELINE.template.md) so every worker reads the same version.
Measure
Track response rate (any reply that is not silence) by role type, source and base resume. Shift the Searcher's mix toward what gets replies.
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
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.

