# Network Jobs

Find jobs where you're already connected. Local-first agent skill suite (MIT,
free): your LinkedIn graph becomes a private job corpus on your machine.
No cloud, no account.

- Landing: https://hirefrank.com/network-jobs/
- Repo: https://github.com/hirefrank/network-jobs
- npm: https://www.npmjs.com/package/@hirefrank/network-jobs

## Setup (recommended): paste into any agent

Set up Network Jobs on this machine: if the network-jobs CLI isn't installed,
run npm i -g @hirefrank/network-jobs@latest then network-jobs setup --agent
auto. Then follow
https://github.com/hirefrank/network-jobs/blob/main/skills/network-jobs-setup/SKILL.md
- ask me for my resume and LinkedIn Connections export, import both, and run
the preferences interview.

## Setup (terminal alternative)

npm i -g @hirefrank/network-jobs@latest && network-jobs setup --agent auto

Works with Claude Code, Cursor, Codex, OpenCode, Muse, Pi, Gemini CLI,
Hermes, Grok Bot, and any agent that reads skills and runs local commands.
`setup --agent auto` links every detected agent (Muse included, via the
shared `~/.agents/skills`); long-running agents resume from local files.

## The loop (in the agent)

1. `network-jobs import <zip-or-csv>` - LinkedIn Connections export becomes
   `connections/` + `companies/` under `~/.network-jobs/`.
2. `careers-discover` - model-driven career-page discovery. Stages under
   `triage/` only. Never ingests on its own.
3. Human reviews triage and names the batch to take.
4. `jobs-ingest` - confirmed triage batch becomes normalized `corpus/` shards.
   Only on explicit confirmation.
5. `network-jobs search` - prefs-aware ranked search -> `search/ranked.json`
   (`matchScore` + `matchReasons` per job).
6. `intro-email-generator` - forwardable warm intro through a real mutual ->
   `search/intros.json`. The agent drafts; the human sends.

## Skills (6)

- `network-jobs-setup` - resume -> profile.json + preferences interview.
- `network-jobs-import` - Connections ZIP -> local company graph.
- `careers-discover` - career pages -> triage batches.
- `jobs-ingest` - confirmed batches -> searchable corpus.
- `network-jobs` - prefs-aware corpus search.
- `intro-email-generator` - warm intros via real connections.

## Rules

- Data lives under `~/.network-jobs/` (`NETWORK_JOBS_HOME` overrides).
- Never invent job fields, emails, or connections.
- Never send profile, resume, or connections to any remote API.
- Read `SCHEMA.md` in the repo before writing data files.