AI Job Search Profile Setup: Three Onboarding Paths

Before /scrape can find relevant jobs and /apply can generate tailored applications, the framework needs to know who you are. The /setup command collects your professional history, skills, behavioral profile, and job search preferences, then writes them into a set of profile files that every other command reads. This page covers all three setup paths in detail, the files that get populated, and how to update specific sections later. Run /setup from inside a Claude Code session started at the repository root:

claude

/setup

Claude scans your documents/ folder automatically and presents the three paths based on what it finds.

The three setup paths

Path A: Documents folder
Path B: Single CV import
Path C: Interview mode

Path A is the recommended starting point if you have any career documents available. Claude reads everything in the documents/ folder, cross-references the files for consistency, and merges the extracted data into your profile. It is idempotent: re-running it as you add more documents will never silently overwrite existing content. Conflicts are surfaced explicitly for your decision.

What to put in each subfolder

cv/: your master CV

Place your most comprehensive, unedited CV here as a .pdf or .tex file. This is your full professional record, not a tailored variant. Claude extracts work experience (titles, companies, dates, bullets), education (degrees, institutions, dates, thesis topics), technical skills, awards, publications, and contact information. If multiple files are present, Claude reads all of them and cross-references for consistency.Keep tailored CVs out of this folder. The skill files become the canonical source of record, and /apply generates tailored variants per application from them.

linkedin/: LinkedIn profile export

Export your LinkedIn profile as a PDF (Profile → More → Save to PDF) and place it here. Claude extracts work experience and dates (cross-referenced against your CV), skills and endorsements, education, certifications, volunteer work, publications, and the full text of your About section. The About section is used to infer behavioral profile additions, and any recommendations received are used to enrich reference context.If multiple exports are present, Claude uses the most recently modified file.

diplomas/: degree certificates and transcripts

Place PDF degree certificates and official transcripts here. Claude extracts official degree titles and names (used to verify education entries), graduation dates, grades or distinctions if visible, and the official institution name spelling. Descriptive filenames like msc_physics_ucph_2025.pdf help with readability but do not affect parsing.

references/: reference letters

Place reference letters here in .pdf, .txt, or .md format. Claude extracts the referee’s name, title, and organization; specific quotes and assessments (added to 01-candidate-profile.md); and competency language used by referees, which adds behavioral signal to 02-behavioral-profile.md. Use filenames like reference_jane_smith.pdf.

applications/: past application records

Each past application goes in its own subfolder named <company>_<role> (lowercase, underscores for spaces). For example: acme_ml_engineer/ or bigcorp_software_engineer/.Each subfolder can contain four files:

job_posting.md: the original job posting pasted as text. Used to infer which skills and role types you have targeted and to calibrate 04-job-evaluation.md.
cover_letter.tex: the cover letter you submitted. Used to extract writing style patterns and opening structures for 06-cover-letter-templates.md.
cv_draft.tex: the CV variant you submitted. Used to extract profile statement styles for 05-cv-templates.md.
outcome.md: fill this in after the application resolves. Records status (hired, rejected, no response, interview only), interview stages reached, and any feedback you received.

Claude learns from outcomes: applications that reached interview stage are flagged as confirmed strong-fit signals, and patterns across no-response or rejection outcomes inform the calibration in 04-job-evaluation.md.

Cross-reference checking

Before proposing any changes, Claude checks for inconsistencies across documents: date mismatches between your CV and LinkedIn, title mismatches for the same role, education discrepancies. Any inconsistencies found are presented as a numbered list before writing begins, and Claude waits for you to resolve each one.

Behavioral and writing-style additions inferred from documents (rather than stated directly) are labeled as inferred. For example: [Inferred from LinkedIn About, review before relying on this]. This lets you review AI-synthesized additions critically before the profile is relied on for applications.

Path B is for sharing one CV or resume and letting Claude fill in the gaps. You can either paste the CV text directly into the chat or use the @ file-mention syntax to point Claude at the file:

/setup
# Then at the path selection prompt, choose Path B
# and @-mention your file or paste the text

Claude reads the document thoroughly and extracts all structured information: name, contact details, education, work experience, skills, publications, and awards. It then presents a summary of what was extracted and asks follow-up questions for anything missing, including behavioral profile, career goals, deal-breakers, and salary expectations.Path B converges on the same profile files as Path A and Path C. The difference is source richness: a single CV typically captures less context than a full documents folder, so expect more follow-up questions.

Path C walks you through nine structured sections conversationally. Claude asks questions naturally and you answer in your own words. This path is the right choice if you are starting from scratch without existing documents.The sections are:

Identity and contact: name, location, phone, email, LinkedIn, GitHub, languages, employment status
Education: degrees, institutions, dates, thesis topics, certifications
Professional experience: roles, companies, dates, responsibilities, achievements, tools used
Technical skills: programming languages, ML/AI frameworks, domain expertise, software platforms
Publications and awards: peer-reviewed papers, hackathons, competitions (optional)
Behavioral profile: formal assessment results (PI, DISC, Myers-Briggs) or synthesized answers to behavioral questions
Career goals and preferences: target roles, what energizes you, deal-breakers, salary expectations, location constraints
References: names, titles, organizations, relationship context (optional)
Job search configuration: role titles to search for, key skills as search terms, target companies, geographic scope, and portal selection

Section 9 is where Claude generates the search queries that power /scrape. Claude will also proactively suggest role types you may not have considered based on your skill profile. For example, if you have strong domain expertise combined with Python skills, Claude might surface “Technical Consultant” or “Solutions Engineer” as adjacent paths worth searching.

You can skip optional sections (publications, awards, references, behavioral profile) by saying so. The remaining sections are required to generate a usable profile.

Files that get populated

All three paths write to the same core set of files. After a complete /setup run, every file in this table will be populated with your actual information. Files marked “Path A only” are populated when Path A has source material to draw from.

File	What it contains
`CLAUDE.md`	Your full candidate profile, the top-level context file that every command reads first
`.claude/skills/job-application-assistant/01-candidate-profile.md`	Structured education, work experience, independent projects, technical skills, publications, awards, and references
`.claude/skills/job-application-assistant/02-behavioral-profile.md`	Behavioral assessment results or synthesized behavioral traits
`.claude/skills/job-application-assistant/03-writing-style.md`	Tone, structure, and writing-style patterns extracted from past cover letters (Path A only; labeled as inferred)
`.claude/skills/job-application-assistant/04-job-evaluation.md`	Personalized skill match areas, career goals, and motivation filters used to score job postings
`.claude/skills/job-application-assistant/05-cv-templates.md`	Profile statement templates for different role types, drawn from your background
`.claude/skills/job-application-assistant/07-interview-prep.md`	STAR examples from your actual experience (Path A leaves stubs to complete manually; Paths B and C generate full examples)
`cv/main_example.tex`	Your LaTeX CV with actual name, contact details, and initial experience entries
`.claude/skills/job-scraper/search-queries.md`	Job search queries for `/scrape`, organized by priority category and role direction

02-behavioral-profile.md and 03-writing-style.md are populated differently in Path A: behavioral additions are inferred from your LinkedIn About section and reference letters, and writing-style patterns are extracted from past cover letters in documents/applications/. Both are labeled as inferred and not treated as authoritative until you review them.

Updating specific sections

You do not need to re-run the full setup to update one section. Use the --section flag to target any named section from Path C:

/setup --section skills
/setup --section experience
/setup --section search

When you pass --section <name>, Claude skips the path-selection prompt and goes directly to that section in interview mode for an update-only flow.

Updating your job search configuration

The --section search option is the most commonly used after initial setup. As your priorities evolve or you finish a job search cycle and start another, your target roles and geographic scope may change:

/setup --section search

This re-runs the full search configuration interview: which role titles to search for, which skills to use as search terms, which locations to include, and which portals to query. Claude also suggests role types you may not have considered based on your current profile. The result is an updated search-queries.md that takes effect the next time you run /scrape.

Re-running /setup as your profile grows

The documents-folder path (Path A) is designed to be re-run any time you add new material. Each run:

Reads the current state of all profile skill files
Compares newly extracted document content against what is already there
Proposes only content that is genuinely new or conflicting
Never silently overwrites. Conflicts are presented explicitly for your decision

Good times to re-run:

After adding a new LinkedIn export
After receiving and adding a reference letter
After recording outcomes for completed applications in documents/applications/
After updating your master CV with a new role

If you want to start completely fresh, use /reset profile to clear the skill files while preserving the framework rules, or /reset all to also clear your documents folder. Both commands show exactly what will be deleted and require you to type RESET to confirm before anything is removed.

Get Started

Core Workflow

Commands

Customization

Reference

AI Job Search Profile Setup: Three Onboarding Paths

The three setup paths

What to put in each subfolder

Cross-reference checking

Files that get populated

Updating specific sections

Updating your job search configuration

Re-running /setup as your profile grows

Build docs developers (and LLMs) love

Get Started

Core Workflow

Commands

Customization

Reference

Documentation Index

​The three setup paths

​What to put in each subfolder

​Cross-reference checking

​Files that get populated

​Updating specific sections

​Updating your job search configuration

​Re-running /setup as your profile grows

Build docs developers (and LLMs) love

The three setup paths

What to put in each subfolder

Cross-reference checking

Files that get populated

Updating specific sections

Updating your job search configuration

Re-running /setup as your profile grows