Call this FIRST whenever a user wants a resume built, tailored or checked against a job. Returns the step-by-step playbook, the intake checklist (what to collect from the candidate, including the LinkedIn profile), and ATS formatting rules.

Validates that all primary candidate details are collected (name, email, phone, location, LinkedIn URL, current resume) and cross-checks the resume against the LinkedIn profile text. Returns ready=true only when resume generation may proceed.

Decodes a job description into weighted keywords (must-have vs nice-to-have, with exact ATS phrasing), ranked focus points, seniority, years required, education/certifications, domain terms, action verbs to mirror, and implied expectations.

Compares the candidate (resume, LinkedIn, extra context) to the JD. For every missing requirement, assigns a strategy (surface, reframe, ask_candidate, bridge_project, quick_learn). Also maps each JD focus point to the candidate's best existing evidence, proposes JD-themed bridge projects, suggests quick-win certifications, and returns a resume plan: headline, section order, summary formula, a drop-in Skills section with a 'Familiar with' tier, job-title translations (e.g. 'Member of Technical Staff' → 'Software Engineer (MTS)'), and buried experience worth promoting (internships, freelance, open source). Also returns questions to ask the candidate.

Designs realistic, company-themed portfolio projects that close specific skill gaps in days. Each comes with a stack, a day-by-day build plan, resume bullet templates (fill with real measured numbers), and interview talking points. Use it for more or different project ideas than analyze_gaps returned.

Scores a resume 0–100 against a JD the way keyword-ranking ATS and recruiters do: weighted keyword match, keywords backed by bullets, title alignment, bullet quality (verbs, metrics), structure/parseability, length, and integrity (placeholders, stuffing). Pass candidate_sources to get a truth_check of unverified claims. Accepts a structured resume or plain text.

Renders the approved resume to ATS-safe DOCX / PDF / Markdown / TXT. Requires candidate_approved=true and a valid LinkedIn URL. Refuses if placeholder metrics like [N] remain. Also returns LinkedIn alignment suggestions and, if jd_text is given, the final ATS score.

Everything needed to actually apply, generated from the JD and the candidate's materials: an apply-today checklist with portal-specific tips (Workday, Greenhouse, Lever…), an auto-reject check (years, degree, sponsorship, location, must-haves), copy-paste answers to application form questions (including honest 'years with X' computed from role dates), a cover letter, referral and recruiter messages with LinkedIn search links, a LinkedIn connection note, follow-up and thank-you emails, LinkedIn headline/About/skills updates, interview prep (elevator pitch, likely questions, 'defend every bullet'), and a tracker entry. Call it after render_resume, passing the final resume.

Generates a runnable starter repository for a bridge project: a Go, Python (FastAPI) or TypeScript service with a /health endpoint and a passing test, CI, a docker-compose for the stack's databases and brokers, a README carrying the build plan, results table and resume-bullet templates, and a BUILD_LOG. Java gets a Spring Initializr command. Pass a project from analyze_gaps or suggest_bridge_projects.