job-tracker

name	job-tracker
description	Update application status, log follow-ups, and maintain the job tracker spreadsheet. Use this skill whenever the user wants to update the status of a job application — "I applied," "I got a rejection," "they scheduled an interview," "mark it as withdrawn," "update the tracker," "log this application," "I submitted it," "any follow-ups due?" Also trigger when the user says "I'm done" or "submitted" after the resume and cover letter are ready, or asks to check on pending applications, review follow-up dates, or get a status overview of all active applications. This is the bookkeeping skill — it keeps metadata.json and tracker.xlsx in sync so nothing falls through the cracks. Do NOT trigger for evaluating new jobs (job-intake), tailoring resumes (resume-tailor), or writing cover letters (cover-letter-writer).
recommended_model	{"default":"haiku","reasoning":"This skill does pure file operations: read metadata.json, update a field, write it back, update a spreadsheet row. No reasoning, no creative writing, no analysis. Haiku handles this at a fraction of the cost of Sonnet.\n","upgrade_to_sonnet_when":"The user asks for a status summary or analysis across multiple applications (e.g., \"how's my job search going?\"). Summarizing patterns across many applications benefits from Sonnet's reasoning. For individual status updates, Haiku is sufficient.\n"}

Job Tracker — Keep Applications Organized

Intent

1. metadata.json is the single source of truth — tracker.xlsx is a convenience view; when they conflict, metadata.json wins and the spreadsheet gets corrected
2. Follow-up discipline prevents lost opportunities — every "applied" status gets a follow-up date; every passed follow-up date gets surfaced; no application silently expires
3. Folder completeness before status advancement — warn when advancing to "applied" if resume or cover letter files are missing from the application folder
4. No lost applications — every status change is persistent across both metadata.json and tracker.xlsx; nothing falls through the cracks between sessions
5. Searchable history enables learning — past applications are data for analytics and strategy, not dead folders; the tracker is the index that makes them findable
6. Single status update under 1 minute; full overview under 5 minutes — bookkeeping should be invisible, never a bottleneck in the application workflow

Why this skill exists

Applying to jobs without tracking means losing context. Three weeks after submitting, you forget which version of your resume you sent, whether you had a contact name, or when you planned to follow up. This skill makes every status change persistent — updating both the application's metadata.json (source of truth) and the tracker.xlsx (convenience view) so neither drifts.

Prerequisites

This skill expects:

• job-applications/tracker.xlsx — the master spreadsheet
• At least one application folder in job-applications/applications/ with a metadata.json file

If tracker.xlsx doesn't exist, offer to create it using the schema from job-intake's references/folder-schema.md.

Workflow

Step 1: Understand the update

The user will tell you what happened. Common triggers:

User says	Action
"I applied" / "submitted" / "sent it"	Set status → applied, set applied_date to today
"Got a rejection" / "they passed"	Set status → rejected
"They want an interview" / "scheduled a call"	Set status → interviewing
"Got an offer"	Set status → offered
"I'm pulling out" / "withdrawing"	Set status → withdrawn
"Follow up next week" / "remind me on March 5"	Set follow_up_date
"The recruiter's name is Sarah"	Set contact field
"Add a note: they want references"	Append to notes field

If the user doesn't specify which application, check context — did they just finish tailoring a resume or writing a cover letter? If so, use that application. If it's ambiguous, list the active applications and ask which one.

Step 2: Update metadata.json

Read the application's metadata.json, make the change, and write it back.

Status transitions follow this lifecycle:

                          PIPELINE PHASE                          APPLY/LEARN PHASE
ingested → sourced → pending_review → evaluating → ready_to_apply → applied → interviewing → offered
                           ↘ skipped                                    ↘ rejected
              ↘ unresolved                                     ↘ withdrawn

Pipeline transitions (pending_review → evaluating and pending_review → skipped) are handled by the email-triage skill, not this skill.

Validate the transition makes sense. If someone says "I got an offer" but the status is still evaluating, that's unusual — confirm before updating. The user might mean a different application.

When setting status to applied:

• Set applied_date to today's date (YYYY-MM-DD) unless the user specifies a different date
• Set a default follow_up_date two weeks out unless the user specifies one
• Confirm the resume and cover letter filenames are populated — if they're null, warn the user that the application folder doesn't have tailored materials saved

When setting status to interviewing:

• If the user mentions a date, set follow_up_date to the day after the interview
• If they mention a name, update contact

Step 3: Update tracker.xlsx

After updating metadata.json, sync the change to tracker.xlsx. The spreadsheet is a convenience view — metadata.json is the source of truth, and the spreadsheet should reflect it.

Read the existing tracker.xlsx. Find the row matching this application by company and role. Update the relevant columns:

Column	Source
Date Applied	metadata.applied_date
Company	metadata.company
Role	metadata.role
Source	metadata.source
Status	metadata.status
Match Score	metadata.match_score.overall
Follow-up Date	metadata.follow_up_date
Contact	metadata.contact
Resume Version	metadata.resume_version
Cover Letter	metadata.cover_letter
Notes	metadata.notes

If no row exists for this application (e.g., the user applied without going through the full skill chain), add a new row.

Read references/xlsx-operations.md for the specific openpyxl patterns for reading and updating the spreadsheet.

Step 4: Confirm the update

Tell the user what you changed, briefly:

• "Updated HealthFirst VP of Engineering to applied — applied date set to Feb 24, follow-up date set to March 10. Tracker spreadsheet is in sync."

If there are other applications with follow-up dates that have passed, mention them: "By the way, you have 2 applications with follow-up dates that have passed — want me to review them?"

Bulk operations

The user may ask for operations across multiple applications:

"How's my job search going?" / "Give me a status overview" Scan all applications/*/metadata.json files. Summarize by status:

• X applications in evaluating
• X applied, waiting to hear back
• X in interview process
• X with follow-ups due (list them with dates)

"Update the tracker from all folders" Rebuild tracker.xlsx entirely by scanning all metadata.json files. This is useful if the spreadsheet has drifted or the user has been updating metadata manually.

"What follow-ups are due?" Scan for applications where follow_up_date is today or earlier and status is applied or interviewing. List them with company, role, and how many days overdue.

Skill composition

This skill is the final step in the job application workflow:

Upstream	This skill
job-intake → resume-tailor → cover-letter-writer	job-tracker (logs and tracks)

After this skill runs, the application is fully tracked. The workflow is complete until the user gets a response and needs another status update.

Edge cases

• Application not in the system — The user applied outside the skill chain (e.g., quick-applied on LinkedIn). Create a minimal metadata.json and folder from whatever info the user provides, then add to the tracker.
• Duplicate applications — If a folder already exists for this company+role at a different date, ask if this is a re-application or if they meant the existing one.
• Tracker spreadsheet is missing — Offer to create it. Scan all existing application folders to populate it.
• Bulk rejection — The user might say "reject everything from last month that I haven't heard back from." Process each one individually but confirm the list before making changes.