Job Hunt

Local-first job search tooling for discovering roles, evaluating fit, tracking pipeline state, and generating reviewable resume/cover-letter drafts.

The repository is designed to keep reusable tooling in Git while keeping candidate-specific data local.

What This Does

• Scans supported job sources.
• Extracts structured job descriptions.
• Scores jobs against a search profile.
• Tracks application state in a terminal dashboard.
• Builds draft candidate profiles from CV/resume sources.
• Generates draft resume and cover-letter material for human review.

It does not auto-submit applications.

Data Boundary

Do not commit private candidate data.

Ignored by default:

• .env
• applications/
• generated tracker CSV files in data/
• generated profile-builder outputs under profile_build/
• built dashboard binaries

The important contract is documented in DATA_CONTRACT.md.

Layout

skills/                 Python command-line tools
dashboard/              Go Bubble Tea terminal UI
statuses.yaml           Canonical application statuses
DATA_CONTRACT.md        User-data vs system-code boundary
profile_build/README.md Profile-builder output notes
interview-prep/         Story-bank scaffold

Local-only data normally lives in:

resume.yaml
search_profile.yaml
companies.yaml
applications/
data/applications.csv
profile_build/
.env

Requirements

Python:

• Python 3.10+
• PyYAML for YAML-backed commands
• pdftotext for PDF profile ingestion

Go dashboard:

• Go 1.18+

Local model:

• The Python tools expect an OpenAI-compatible local LLM endpoint.
• Configure it through .env or environment variables used by skills/_common.py.

Core Workflow

Run a profile-driven search:

skills/find_jobs.py profile --limit 10

Draft resume and cover-letter material for selected applications:

skills/apply_to.py applications/example-role-YYYY-MM-DD

Render final polished markdown files to PDFs:

skills/render_pdf.py applications/example-role-YYYY-MM-DD

Log an outcome:

skills/log_outcome.py applications/example-role-YYYY-MM-DD --outcome applied --source "Company site"

Dashboard

Build the dashboard:

cd dashboard
go build -o career-dashboard .

Run it:

dashboard/career-dashboard --path .

Useful keys:

j/k       move
h/l       switch status tabs
/         search
s         cycle sort
c         change status
a         run apply_to.py
p         run render_pdf.py
r         refresh
q         quit

Profile Builder

Build draft profile artifacts from CV/resume-like sources:

skills/build_profile.py ingest resume.pdf linkedin.html old-cv.md

Outputs are written under profile_build/:

profile.json
resume.draft.yaml
search_profile.draft.yaml
evidence_catalog.yaml
evidence_catalog.md
profile_notes.md

These are review artifacts. Merge only verified facts into real profile files.

The builder is domain-neutral. It uses cheap deterministic extraction for obvious facts, then asks the local model to infer appropriate domains, role buckets, proof points, resume variants, cover-letter angles, and verification items.

Supported Sources

Job source adapters currently include:

• Ashby
• Greenhouse
• Lever
• RemoteOK

Watched company configuration lives in companies.yaml.

Search targeting lives in search_profile.yaml.

Safety Rules

• Review generated material before using it.
• Keep .env, resumes, applications, PDFs, and tracker outputs out of Git.
• Treat profile-builder output as draft evidence, not truth.
• Do not claim skills, credentials, dates, or metrics unless verified.
• Prefer updating the evidence catalog before tuning resume and cover-letter prompts.

Checks

Syntax-check Python tools:

PYTHONDONTWRITEBYTECODE=1 python3 -c 'import ast, pathlib; [ast.parse(pathlib.Path(p).read_text()) for p in pathlib.Path("skills").glob("*.py")]'

Build dashboard:

cd dashboard && go build -o career-dashboard .

Smoke-check dashboard data loading:

dashboard/career-dashboard --check --path .