Supervised Job Application Agent: human-in-the-loop application automation
By Syed Ahmed Basharat Ali. github.com/sa-basharat-ali/job-application-agent
A supervised agent that finds, scores and tailors job applications but stops before final submission, so every sensitive decision stays with the applicant.
Challenge
Most job-search automation optimizes for volume over quality, sending generic applications and taking irreversible actions such as submitting forms or accepting agreements without the applicant in the loop.
Approach
- Built a supervised agent that finds, scores and tailors applications but stops before final submission.
- Packaged the workflow as a reusable Codex skill with explicit safety checkpoints.
- Wrote a dependency-free Python CLI for workspace management, duplicate and exclusion detection, and privacy validation.
- Used private workspace templates to store verified facts, job preferences and application summaries.
Results
- Keyword-based fit scoring for fast triage before any document is generated.
- Resume tailoring that preserves a master DOCX template, with a required PDF inspection step before form preparation.
- Privacy scanning that prevents accidental commits of sensitive data.
- Hard stops at CAPTCHAs, authentication, legal agreements and sensitive declarations.
Technologies
Python 3.10+, Codex skill framework, LibreOffice or Word (DOCX to PDF), optional Chrome integration, MIT licensed.
Impact
It inverts the usual automation trade-off, putting truthful, well-matched applications and applicant control ahead of submission volume.