Help & Getting Started

This is a private, invite-only workspace for tracking job applications and generating tailored résumés and cover letters. It stores your canonical profile data as structured YAML, lets you layer role-specific variants on top, and exports polished documents without any manual copy-paste.

Quizzes and playbooks are shared across all users — they cover interview prep, site optimization, and LinkedIn strategy. Your application data and profile are scoped to your own tree; no other user can see them.

The dashboard is built on Next.js 15 and deployed on Azure App Service behind GitHub OAuth. Source code lives at github.com/Scarcrux/Careers (private).

Everything in the dashboard relates to one of these eight concepts.

Navigate to /variants and click into the variant you want to generate for. On the variant detail page at /variants/{profileId}/{role}, use the Download Résumé and Download Cover Letter buttons to export DOCX files.

Two query parameters let you further tailor the output:

• ?company={slug} — pulls in the company entry overlay for company-specific framing.
• ?template=modern or ?template=classic — switches the DOCX template. Defaults to modern.

Example: /variants/jonathan/qr?company=two-sigma&template=classic

Generated files are not stored in the repo — they are ephemeral. If the output looks wrong, fix the profile or variant data and regenerate; never edit the DOCX directly.

Add an application from /applications by pasting a posting URL, or run the scrape CLI and then promote a scraped job to an application. Each application moves through the following status flow:

interested — saved from the job board but not yet submitted. applied — submitted. interviewing — any stage of the interview process. offer, rejected, and withdrawn are terminal states.

The timeline on each application records every status transition with a date. You can also attach structured feedback (technical screen, hiring manager, etc.) and interview slots.

Open an application at /applications/{id} and scroll to the Interviews section. Add an InterviewSlot with a date, time, duration, and interview kind (phone, technical, behavioral, etc.).

Each slot has a Download .ics button that exports a calendar event compatible with Google Calendar, Apple Calendar, and Outlook. Use this to keep your interview schedule in sync with your personal calendar.

Upcoming interviews surface on the home page and at /calendar, sorted by date.

Both are shared content — every user on the dashboard has access to the same library regardless of their profile or variant setup.

Playbooks at /playbooks are structured markdown guides. Topics include LinkedIn profile optimization, personal site SEO strategy, cover letter structure, and role-specific interview playbooks.

Quizzes at /quiz come in three types:

• Knowledge — multiple-choice questions. Select an answer and submit.
• Math — numeric input. Type your answer; the checker accepts exact values or configurable tolerance.
• Coding — in-browser Python execution via Pyodide. Write your solution and run it; no server round-trip.

The Parse with Claude button on /profiles/new calls Claude Haiku via the Anthropic API. It extracts structured profile data from raw résumé text in a single pass.

Quotas — each user is capped at 10 parses per day and 100 parses per month. Counters reset at midnight. The parser also rejects inputs longer than 14,000 characters — paste just the text content of your résumé, not a raw HTML or PDF dump.

If the parse fails: scroll to the bottom of /profiles/new — there is a full manual entry form as a fallback. You can always fill fields by hand without using any quota.

Parse output is a starting point, not ground truth. Always review and correct before saving, especially dates, GPA, and project descriptions.

Your data lives in a dedicated path on the server: data/users/{your-github-login}/. Application data, profile data, parse-usage counters, and generated artifacts are all scoped to that path. No other user can read or write to it — access checks are enforced server-side on every request.

Quizzes, playbooks, and scraped job listings are shared and read-only for all users.

Admins can view the system-wide allowlist (data/allowlist.yml) and parser configuration (data/parse-config.yml), but cannot access individual user data directories through the dashboard UI.

Authentication is handled by GitHub OAuth — the dashboard requests identity scope only and stores no GitHub tokens beyond the session. Sessions are signed with AUTH_SECRET and expire after inactivity.

This is a small, invite-only tool — there is no support ticket queue or public forum. Ping the person who invited you directly. If something is clearly broken (parse returning 500, downloads failing, calendar not rendering), include the URL you were on and what you were trying to do.

For feature requests or data corrections, the source of truth is the YAML in the repo. Admins can push fixes directly to the Azure Files mount and they take effect immediately without a redeploy.