# AI Job Search /setup Onboarding Paths: Three Ways to Initialize Your Profile

> Explore the three /setup onboarding paths in AI Job Search: Documents scan, CV import, and Interview mode. Initialize your profile efficiently with MadsLorentzen/ai-job-search.

- Repository: [Mads Lorentzen/ai-job-search](https://github.com/MadsLorentzen/ai-job-search)
- Tags: getting-started
- Published: 2026-08-05

---

**The `/setup` command in the ai-job-search framework offers three onboarding paths—Documents folder scan, Single CV import, and Interview mode—all converging on the same seven profile files.**

When you run `/setup` in the MadsLorentzen/ai-job-search repository, Claude evaluates your repository state and presents **three distinct onboarding paths** designed to accommodate different starting points. These paths are defined in [`.claude/commands/setup.md`](https://github.com/MadsLorentzen/ai-job-search/blob/main/.claude/commands/setup.md) and documented in [`SETUP.md`](https://github.com/MadsLorentzen/ai-job-search/blob/main/SETUP.md), ensuring consistent downstream behavior for commands like `/apply` and `/scrape`.

## The Three /setup Onboarding Paths Explained

### Path A: Documents Folder Import (Recommended)

Use this path when your repository already contains files in the `documents/` directory—CVs, LinkedIn exports, diplomas, references, or other career artifacts.

Claude performs a **read-before-write scan** of the entire `documents/` tree, cross-references data across files, and automatically populates:

- Seven skill files ([`01-candidate-profile.md`](https://github.com/MadsLorentzen/ai-job-search/blob/main/01-candidate-profile.md), [`02-behavioral-profile.md`](https://github.com/MadsLorentzen/ai-job-search/blob/main/02-behavioral-profile.md), etc.)
- Generated LaTeX CV (`cv/main.tex` or similar)
- Search query files ([`.claude/skills/job-scraper/search-queries.md`](https://github.com/MadsLorentzen/ai-job-search/blob/main/.claude/skills/job-scraper/search-queries.md))

This process is **idempotent**—you can rerun `/setup` as you add more documents without duplicating data.

```bash
claude
/setup
> go  # confirms Path A when documents/ has files

```

### Path B: Single CV Import

Select this path when you want to start from one résumé—uploaded directly or @‑mentioned in the conversation.

Claude extracts structured fields:

- Name and contact information
- Education history
- Work experience
- Technical skills
- Languages
- Awards and certifications

For any missing data, Claude asks follow‑up questions before generating the same profile files as Path A.

```bash
claude
/setup
> @my_resume.pdf  # or paste CV text directly

```

### Path C: Interview Mode

Choose this path for conversational onboarding when you have no documents, or prefer natural‑language interaction.

Claude walks you through structured sections:

1. Identity and contact details
2. Education background
3. Work experience
4. Technical skills and stack
5. Behavioral profile
6. Career goals and preferences
7. Job-search configuration (role titles, geographic scope, portal preferences)

Answers are synthesized into the profile files, with additional search parameters captured for the scraper.

```bash
claude
/setup
> interview

```

## How the Paths Converge

All three `/setup onboarding paths` resolve to identical outcomes:

| Step | Action |
|------|--------|
| **Step 3** | Profile file generation in `.claude/skills/job-application-assistant/` |
| **Step 4** | Confirmation and validation |

This convergence guarantees that downstream commands—`/apply`, `/scrape`, `/letter`, `/interview`—operate on a fully populated, consistent data structure regardless of how you onboarded.

## Implementation Details

The onboarding logic lives in two key files:

- **[`SETUP.md`](https://github.com/MadsLorentzen/ai-job-search/blob/main/SETUP.md)** (lines 102-107): High‑level overview listing the three paths
- **[`.claude/commands/setup.md`](https://github.com/MadsLorentzen/ai-job-search/blob/main/.claude/commands/setup.md)** (lines 17-30): Command implementation with welcome message construction and detailed path handling

The welcome message dynamically adapts based on repository state. If `documents/` contains files, Path A is pre‑selected and highlighted; otherwise, the three options appear as equal choices.

## Summary

- **Three `/setup onboarding paths`** serve different user starting points: existing documents folder, single CV, or conversational interview.
- **Path A** (Documents folder) is fully automated and idempotent—ideal for users with prepared materials.
- **Path B** (Single CV) balances automation with structured follow‑up for incomplete data.
- **Path C** (Interview mode) captures job‑search configuration parameters that Paths A and B infer from documents.
- All paths output identical profile files to `.claude/skills/job-application-assistant/`, ensuring command interoperability.

## Frequently Asked Questions

### What happens if I rerun `/setup` after adding new documents to the folder?

Claude's Path A implementation is **idempotent**. Re-running `/setup` re‑scans `documents/`, cross‑references against existing profile data, and updates files without duplication. Your previous answers and generated search queries are preserved unless superseded by new document content.

### Can I switch paths mid‑onboarding if I change my mind?

The [`setup.md`](https://github.com/MadsLorentzen/ai-job-search/blob/main/setup.md) command structure presents the three paths as an initial choice. If you select Path C (Interview) but then paste a CV, Claude detects document content and can pivot to extraction behavior. For explicit path changes, restart with `/setup` and select your preferred option.

### Does Interview mode (Path C) collect different data than the document paths?

Path C explicitly captures **job‑search configuration**—role titles, geographic scope, and portal preferences—that Paths A and B infer from document analysis. All paths populate the same seven skill files, but Path C's conversational flow ensures these search parameters are defined even without document metadata.