Skip to content

CV and applying

github-actions[bot] edited this page Nov 17, 2026 · 1 revision

Your CV and applying

Your CV, written once, and everything written from it for each job: a CV tailored to the posting, a cover letter, messages to the people hiring, interview prep from your own stories. The assistant writes them from what your CV says, never more; you review and send.

CV

CV (/cv): upload your CV (PDF, Word or text, up to 5 MB), kept in PostgreSQL with the text read out of it (PDFBox for PDFs), or fill in the fields (personal details, summary, experience, projects, education, skills, languages, certifications) and JobWarden builds one: previewed as HTML, downloaded as a PDF made from the same template with openhtmltopdf. The PDF embeds Noto Sans (SIL Open Font License, in web/src/main/resources/static/fonts), so Romanian, Polish and German letters show. The preview has two tabs: the CV built from the fields, and your file, which the browser shows as it is when it's a PDF, and JobWarden shows as HTML when it's a Word document (its paragraphs, headings, bold and italic text, lists, tables, links and text boxes, read with StAX, every text escaped; DocxHtml) or as its text. The assistant reads the CV with the MCP tool get_cv.

Projects from GitHub

Projects from GitHub: on the CV page, connect your GitHub account by user name (the public repositories), or with a fine-grained token, read-only, for the repositories you pick (the private ones too: what the token is given is what the assistant may read). Tick up to 12 repositories, then ask the assistant to write your CV's projects from them (a WRITE_PROJECTS request): list_github_projects gives the assistant each one's description, languages, topics, stars, dates and the start of its README, which JobWarden reads from GitHub's API with your token; the assistant writes up to six projects from them, never claiming what a repository doesn't show, and saves them with save_cv_projects, in place of the ones you had. They show among the fields to review, and in the CV and its PDF. The token is kept encrypted, like the mailboxes' passwords (github_accounts, in cv).

AI-proof CV

AI-proof CV: on a job's page (and its side panel), the assistant rewrites your CV for that job, so it gets through the screening software employers run CVs through and then reads well to the recruiter. The card asks the assistant with a request, like the AI Assistant's presets; the assistant reads your CV and the posting, puts what you really did in the posting's own words (never a skill, job, date or number you don't have), and saves it with save_tailored_cv, with the posting's keywords it covers, those it asks for that you can't honestly claim, and what it changed. The card shows all three, previews the CV and downloads it as a PDF ("CV Ana Pop - Acme.pdf"), laid out in one column with standard sections. The contact details stay the CV page's. Each job keeps one (tailored_cvs); the CV page lists them.

Cover letter

Cover letter: on a job's page (and its side panel), the assistant writes a cover letter for the job (a WRITE_COVER_LETTER request): it reads the posting and your CV, and saves a letter of four short paragraphs with save_cover_letter, in the posting's language, from the greeting (the job's contact by name when it has one) to the sign-off, with what it highlighted and what to check. Change it on the card (or write one without the assistant), copy it, preview it, or download it as a PDF laid out like the CV, with the CV page's name and contact details at the top, the date and the job ("Cover letter Ana Pop - Acme.pdf"). Each job keeps one (cover_letters, in cv). Auto-apply uses it when a form asks for one.

The Cover letters page

Cover letters (/letters): every job's cover letter and messages in one place. On top, your base letter: written once, in your voice, with the paragraphs you always use, and {company} and {role} where the job's company and title go (base_letters, in cv). A letter you write yourself starts from it, filled in for the job, and the assistant keeps its voice and structure when it writes one (get_request gives it as base_letter, and get_cover_letter for a job without a letter). Below, the jobs in use, those still to apply to first, each with its letter or none, the assistant writing it, and how many people its messages are to; jobs that ended show only when they have a letter. Pick one and its letter and messages open beside the list, the same cards as on its page. Tick the jobs without a letter and ask the assistant for them at once, up to 10: a WRITE_COVER_LETTER request each, and one prompt that does them one after the other.

Reach out

Reach out: also on a job's page, the assistant writes to the people who hire for it (a WRITE_OUTREACH request): its contact, the recruiters on the Who's hiring page who are its contact or work at its company, and the authors of the posts it was saved from or that say its company is hiring, which the request lists for it; with nobody known, it looks for the recruiter with two searches at most and adds them to Who's hiring. For each person, save_outreach_draft keeps a LinkedIn connection note, 300 characters at most (counted as LinkedIn does, and refused when longer), and an email. The card shows them to change, with the note's count, buttons that copy what's in the fields, a link to the person's profile and one that opens the email in your own email app. You send them yourself: JobWarden never sends anything, and the MCP server tells the assistant never to (outreach_drafts, in cv).

Story bank

Story bank: on the CV page, your stories for interviews, told the STAR way (the situation, the task, what you did, the result), with what each shows (skills), up to 40 (cv_stories, in cv). Write them there, or ask the assistant to draft them from your CV (a WRITE_STORIES request, save_stories, up to eight at once): only what the CV says, with a question in brackets where it doesn't say, e.g. "[How many users did it reach?]"; those are marked to finish, and those whose title the bank has are left out. The interview briefs draw on them: the assistant reads them with list_stories, tells them first, and names those that answer what they'll look for (story_ids); the brief shows them, each unfolding to its parts.

Interview prep

Interview prep: on a job's page (and its side panel), the assistant prepares you for its interview (a PREPARE_INTERVIEW request): it reads the posting and your CV, finds out about the company from a few pages (its own, the news, interview reviews), and saves a brief with save_interview_brief to read the evening before: who they are and what's new with them, what the role is about and what they'll look for, the questions likely asked, your own stories that answer them (from your story bank first, else from your CV, never what neither supports), questions to ask them, what to look into before saying yes, and where it read it all (interview_briefs, in jobs).

At a glance

At a glance: under a job's facts, From the AI Assistant shows its fit (the score, the verdict, and, unfolded, why) and each card the assistant makes for it, the AI-proof CV, the cover letter, the messages, the interview prep and at Offer the negotiation, as generated, in progress or not yet: a click goes to the card, and unfolds it (JobAssistantController).

Auto-apply

Auto-apply (/apply): the assistant applies to your Saved jobs for you, in your own Chrome with your own sign-ins (Claude Code with Claude in Chrome: claude --chrome). You tick the jobs, choose whether the assistant sends each form or fills it in for you to check and send, and paste the prompt the page gives into Claude Code (an AUTO_APPLY request). For each job the assistant tailors your CV to it if it isn't yet, downloads it as a PDF to upload (from /claude/cv?profile=…&job=…, which, like the MCP server, trusts whoever can reach the server), and fills the form in from your CV's contact details and your answers, kept on the page: where you can work, visa sponsorship, notice period, earliest start, salary expectation, relocation, where and how you want to work, years of experience, cover letters (the assistant uses the job's own when it has one, and otherwise writes one per job, or only when a form needs one, and keeps it on the job's page so you see what was sent; the PDF to upload is at /claude/cover-letter?profile=…&job=…), the voluntary questions (declined by default), and your own questions and answers. It never makes an answer up: a question none of them covers stops the job, and the page asks it to you; your answer is kept for the next forms, and the job tries again. The assistant skips LinkedIn Easy Apply and sites whose terms forbid automated applying, and stops at a CAPTCHA for you to solve. The page shows how each job went (applied, filled in for you to send, waiting for your answer, skipped, failed); a job the assistant applied to moves to Applied, with what it did on its timeline, and shows by the AI Assistant next to its status in the list, the pipeline and its page (applications.applied_by; the apply module: application_answers and apply_items).

Clone this wiki locally