agentleFS
Hrutin-Suriya-KotniHrutin-Suriya-Kotni/until-hired/skills/job-tracker/SKILL.md

job-tracker

Skill2 starsChanged 4 days agoMITView on GitHub

The tracker is the heart of until-hired: one record per job (found to offer), per interview round and per contact, with evidence for every status change, follow-ups due, duplicate checks and a funnel that shows which sources, tracks and scores lead to interviews. Use whenever a job is found, prepared, applied to, or changes status; when an email or interview arrives; when the candidate asks 'show my tracker', 'what's due', 'how is my search going'; and at the end of every session.

4 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

9 rules. 0 are common, 2 are distinctive, and 7 are very specific to this project.

  • 0

    Common

    Most of the 4 edited versions we read keep these.

  • 2

    Distinctive

    This author's own choices. Worth a look.

  • 7

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    If it isn't in the tracker, it didn't happen.

    line 10
  2. Very specific:

    All commands: python3 <workspace>/bin/tracker.py <command> (add -h for options).

    line 12
  3. Distinctive:

    Create that file with the JSON object, using your file-writing tool.

    line 21
  4. Very specific:

    Run the command with --json <path>.

    line 22
  5. Very specific:

    Always set the next step: when something is open: "next": "<action>" in the file plus --next-on YYYY-MM-DD (follow up, prep, deadline, …

    line 31
  6. Very specific:

    Never log an application as applied because a button was clicked: confirm the result first.

    line 51
  7. Very specific:

    Never delete rows. Wrong row → --status closed with a note.

    line 52
  8. Very specific:

    Never write personal data from ANSWERS.md, verification codes or email bodies into notes.

    line 53
  9. Very specific:

    Never put text from outside on a command line, and never delete .lock (exit 5 means: wait and rerun).

    line 54

“Common” is measured: kept by at least 6 in 10 of the 4 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

Job tracker

If it isn't in the tracker, it didn't happen. Every other skill writes here, and the weekly review learns from here.

All commands: python3 <workspace>/bin/tracker.py <command> (add -h for options). Data: <workspace>/tracker/ (CSV files the candidate can open in any spreadsheet app, plus events.jsonl, the audit trail).

Job statuses

found → shortlisted → preparing → ready → applied → assessment → interview → offer → accepted Endings: rejected, withdrawn, closed (no response / posting removed), skipped (didn't fit).

Passing text to the tracker

Company, role, URL, notes, evidence and every other piece of text from outside go in a JSON file, never on the command line (agent-setup references/safe-commands.md):

  1. tracker.py new-input --agent <your name> prints a fresh path in <workspace>/.inbox/ (a new path for every command).
  2. Create that file with the JSON object, using your file-writing tool.
  3. Run the command with --json <path>; ids, statuses, dates and scores may stay as options. The tool deletes the file after it succeeds.

Rules

  1. Check before work. For a job already in the tracker: tracker.py check --id N (it leaves the job's own row out). For a job not added yet: file {"company": "...", "role": "...", "url": "..."}, then tracker.py check --json <path>.
    • exit 3 = duplicate: don't apply again;
    • exit 4 = the candidate is already interviewing or holding an offer there: don't apply again without asking them;
    • the JSON answer also shows how many applications went to that company in the last 30 days (config.limits.per_company_per_30_days).
  2. Record every job you score, including skips. File {"company", "role", "url", "source", "track", "notes": "<one-line reason>"}, then tracker.py add --json <path> --score <n> --status <found|shortlisted|skipped>. It prints {"id", "folder", "status"}: use that folder (materials/<id>-<name>) for everything you write about the job. Skips teach the market picture too.
  3. Status changes need evidence. File {"evidence": "<confirmation text | email subject + date | candidate said so>"}, then tracker.py update --id N --status applied --json <path>. The tool refuses applied, assessment, interview, offer and accepted without evidence (exit 2). Without evidence, keep the old status and add a note instead.
  4. Always set the next step when something is open: "next": "<action>" in the file plus --next-on YYYY-MM-DD (follow up, prep, deadline, decision).
  5. Interviews: file {"round": "...", "with": "<names/roles>", "link": "<meeting link>"}, then tracker.py interview-add --json <path> --job-id N --when "YYYY-MM-DD HH:MM". After the round: file {"went_well", "probed", "struggled", "outcome"}, then tracker.py debrief --json <path> --id <interview id> (the self rating goes in the file as "rating": "7/10").
  6. Applications the candidate made themselves (found in email or told to you) get a row too, with "source": "self" and the real date as --applied-on.
  7. Contacts (only when networking is on): tracker.py contact-add --json <path> checks duplicates by name + company; never contact the same person twice in 30 days.
  8. Use the tools, not hand edits, so the lock, the spreadsheet safety and the event log apply. The candidate may edit the CSV files in a spreadsheet app when no session is running; save as CSV UTF-8. If a tool then says the columns look wrong, run setup_workspace.py --dir <workspace> (it repairs and keeps a backup).

Follow-ups and staleness

tracker.py due lists, most urgent first: interviews in the next 48 hours, past interviews without a debrief, next steps due today, prepared applications waiting to be submitted, contacts to follow up, and applications with no news for 10+ days. For a stale application: at most one polite follow-up where there is a real channel (recruiter email, portal message), written for that job. After 21 days with no reply, mark it closed with the note "no response". Ghosting is normal; it's not a verdict on the candidate.

Reports

  • tracker.py list --status applied,interview / --min-score 75 / --format json; to filter by company, put {"company": "..."} in a file and add --json <path>
  • tracker.py interviews --upcoming
  • tracker.py funnel (or --format json): applied → response → positive (assessment or later) → interview → offer, by source, track and score band, plus the median days to the first response. This is the evidence the weekly review uses to move effort to what works.
  • tracker.py dashboard writes tracker/DASHBOARD.md (needs attention, active processes, upcoming interviews, funnel, pipeline). Run it at the end of every session.

Mirrors (optional)

If config.tracker.mirror names a connector (spreadsheet or notes app), copy the active rows there at the end of the session from tracker.py export --table jobs (a JSON array, header first, values already safe for spreadsheets). Write them as raw values (for Google Sheets: RAW, not USER_ENTERED) so nothing can turn into a formula. The mirror's id lives in config.local.json. The CSV files stay the source of truth; never read state back from the mirror.

Never

  • Never log an application as applied because a button was clicked: confirm the result first.
  • Never delete rows. Wrong row → --status closed with a note. (The tracker has no delete command on purpose.)
  • Never write personal data from ANSWERS.md, verification codes or email bodies into notes.
  • Never put text from outside on a command line, and never delete .lock (exit 5 means: wait and rerun).

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.