How to Manage Application Pipeline Tracking with Integrity Checks in Career-Ops

Career-Ops manages application pipeline tracking with integrity checks by storing every job application in a single data/applications.md tracker and enforcing consistency through atomic merge scripts, canonical state normalization, file-based locking, and score-based deduplication.

The santifer/career-ops repository treats your job search as a formal data pipeline. Instead of manually editing spreadsheets, it centralizes application pipeline tracking with integrity checks by routing all updates through validation scripts that prevent duplicates, reject failed batch reports, and normalize every status to a canonical vocabulary.

How the Tracker Pipeline Works

Every job application lives in one master file: data/applications.md. Batch workers never touch this file directly. Instead, they write new rows as TSV files into batch/tracker-additions/. A chain of integrity scripts then ingests, validates, and cleans those additions before they become permanent.

Batch Additions and the Merge Stage

Workers create TSV files such as batch/tracker-additions/023-Acme-Senior-AI-Engineer.tsv. Each file contains a single tab-separated row with columns for report number, date, company, role, status, score, flag, report link, and notes.

The merge-tracker.mjs script processes these files atomically:

node merge-tracker.mjs --dry-run   # preview changes

node merge-tracker.mjs             # apply changes

During a merge, the script performs three critical checks. It normalizes statuses using validateStatus, deduplicates by company plus role plus report number, and rejects any row whose report number appears in batch/batch-state.tsv with a failed status.

Deduplication by Score and Pipeline Status

After a merge, dedup-tracker.mjs removes duplicate rows that describe the same opening:

node dedup-tracker.mjs --dry-run   # preview removals

node dedup-tracker.mjs             # perform deduplication

The script groups rows using normalizeCompany and exact-normalized role strings from normalizeRole. When two rows refer to the same report number, the script keeps the entry with the higher score. If scores are equal, it uses statusRank to preserve the more advanced pipeline status:

function statusRank(status) {
  const map = { skip: 0, discarded: 0, rejected: 1, evaluated: 2,
                applied: 3, responded: 4, interview: 5, offer: 6, hired: 7 };
  return map[normalizeStatus(status)] || 0;
}

Normalizing Canonical Pipeline States

Both scripts enforce a single source of truth for status labels. The validateStatus function in merge-tracker.mjs and the normalizeStatus function in dedup-tracker.mjs map every spelling variant to the canonical states defined in templates/states.yml.

The validateStatus helper strips markdown decorations and trailing dates, then resolves aliases:

function validateStatus(status) {
  const clean = status.replace(/\*\*/g, '').replace(/\s+\d{4}-\d{2}-\d{2}.*$/, '').trim();
  const lower = clean.toLowerCase();
  const aliases = { 'aplicado': 'Applied', 'entrevista': 'Interview' };
  return CANONICAL_STATES.find(s => s.toLowerCase() === lower) ||
         aliases[lower] || 'Evaluated';
}

Any unrecognized status falls back to Evaluated and triggers a warning, preventing malformed labels from polluting the tracker.

Enforcing Concurrency Safety

Concurrent writes are blocked by a file-based lock managed in tracker-utils.mjs. Before any script modifies data/applications.md, it acquires the lock in the tracker-lock/ directory. This guarantees that only one process can mutate the tracker at a time, eliminating race conditions between batch workers and manual CLI operations.

After a successful merge, the source TSV is moved to batch/tracker-additions/merged/ so it cannot be processed again. For targeted single-row updates, use set-status.mjs, which applies the same lock and canonical-state logic.

Running the Complete Integrity Workflow

You can add a new application programmatically and run the full pipeline from the shell:

const fs = require('fs');
const tsv = [
  '024',
  '2026-08-20',
  'Beta Labs',
  'Machine Learning Engineer',
  'Applied',
  '4.8/5',
  '✅',
  '[024](reports/024-beta-ml-engineer-2026-08-20.md)',
  'Sent application via LinkedIn'
].join('\t');
fs.writeFileSync('batch/tracker-additions/024-Beta-ML-Engineer.tsv', tsv);

require('child_process').execSync('node merge-tracker.mjs');

Then run the standard maintenance sequence:

node merge-tracker.mjs
node dedup-tracker.mjs
node verify-pipeline.mjs

The verify-pipeline.mjs script performs a final sanity check for unknown states, missing columns, and formatting drift after the merge and dedup steps complete.

Summary

• Store updates as TSV additions in batch/tracker-additions/ instead of editing data/applications.md by hand.
• Run merge-tracker.mjs to atomically ingest new rows, normalize statuses, and exclude failed reports listed in batch/batch-state.tsv.
• Run dedup-tracker.mjs to collapse duplicate openings, keeping the highest score or most advanced pipeline rank.
• Respect the file lock in tracker-lock/; the scripts acquire it automatically, so never edit the tracker during an active merge.
• Define custom states in templates/states.yml; otherwise validateStatus falls back to Evaluated.
• Schedule verify-pipeline.mjs periodically to catch stray formatting issues before they accumulate.

Frequently Asked Questions

What is the master tracker file in Career-Ops?

The master tracker is data/applications.md. It contains a single markdown table that holds every tracked job application. All writes flow through merge-tracker.mjs or set-status.mjs to preserve table integrity.

How does Career-Ops prevent duplicate job applications?

The system prevents duplicates in two stages. First, merge-tracker.mjs deduplicates by company, role, and report number during ingestion. Second, dedup-tracker.mjs groups rows with normalizeCompany and normalizeRole, then uses statusRank to keep the most favorable entry when collisions remain.

What happens if a batch worker reports a failed status?

If a report number appears in batch/batch-state.tsv with status failed, merge-tracker.mjs rejects the corresponding TSV row entirely. This safety check prevents fabricated or erroneous batch results from entering the tracker.

Can I edit applications.md directly?

No. Direct manual edits bypass the integrity checks, status normalization, and file-based locking system. Always add TSV files to batch/tracker-additions/ and run merge-tracker.mjs, or use set-status.mjs for targeted updates.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}