job-carousel

name	job-carousel
description	Transform a user's resume and target job description into a polished, recruiter-facing LinkedIn career carousel package. Use this when the user wants to create a social media career story, position themselves for a specific role, or highlight their strongest experiences for a target job. This is a fit analysis + strategic positioning engine, not a resume formatter.

Purpose

Transform a user's resume and a target job description into a polished, strategically framed LinkedIn career carousel package. This is not a resume-to-slides converter. It is a fit analysis and strategic positioning engine that produces a premium, recruiter-facing social asset.

When to Use This Skill

Invoke this skill when a user:

• Wants to create a LinkedIn carousel or social media career story
• Has a resume (or raw experience text) and a target role or JD
• Wants to position themselves for a specific job, industry, or audience
• Needs help selecting and framing their strongest experiences for a role

Do not use this skill for general resume editing, cover letter writing, or interview prep (unless the user explicitly asks to extend it there).

Inputs Expected

Required:

1. Resume — file path (.pdf, .txt, .md) or pasted text
2. Job description — file path or pasted text (or at minimum: job title + company + 2–3 sentences about the role)

Optional:

• --mode recruiter | hiring-manager | personal-brand (default: auto-infer)
• --theme clean | dark | bold (default: clean)
• --slides 6 | 7 | 8 | 9 | 10 (default: 7; auto-bumps to 8 when --timeline is used)
• --output <path> (default: output/job-carousel/)
• --export canva — also generate a PPTX file (linkedin_carousel.pptx) for Canva import
• --export figma — also generate SVG files (figma_export/slide_*.svg) for Figma import
• --export canva,figma — generate both
• --timeline — insert a visual career trajectory slide (horizontal timeline with highlighted roles) after the hook slide
• --show-education auto | always | never (default: auto) — controls whether education nodes appear on the timeline slide; auto includes education if the candidate has < 6 years of experience or the degree is relevant to the JD domain

Full Workflow

Step 1 — Locate or Request Inputs

• Check if the user provided resume and JD paths or pasted text.
• If resume is missing: ask "Please share your resume text or file path."
• If JD is missing: ask "Please share the job description or job title + company + key requirements."
• Accept messy inputs — do not require clean formatting.

Step 2 — Parse and Normalize Inputs

Run the pipeline:

npx ts-node src/job-carousel/index.ts \
  --resume <path> --jd <path> \
  [--mode recruiter|hiring-manager|personal-brand] \
  [--theme clean|dark|bold] \
  [--slides 6-10] \
  [--timeline] \
  [--show-education auto|always|never] \
  [--export canva,figma]

This calls resumeParser.ts and jobParser.ts to extract structured CandidateProfile and TargetRoleProfile objects.

From resume, extract:

• Roles, employers, dates
• Achievement bullets (especially those with metrics or ownership language)
• Tools, skills, technologies
• Domains/industries worked in
• Ownership indicators ("led", "owned", "built", "launched", "managed")
• Collaboration/stakeholder language

From JD, extract:

• Job title and seniority level
• Required and preferred skills
• Domain/industry signals
• Leadership and cross-functional expectations
• Recruiter priority signals (what the JD emphasizes most)

Step 3 — Run Fit Analysis

Call fitAnalysis.ts to produce a FitAnalysis object:

• Best fit story: The 1–2 sentence narrative of why this candidate is compelling for this role
• Strongest alignments: Skills/experiences that directly match the JD (score > 7)
• Moderate alignments: Partial matches, adjacent skills, transferable experience (score 4–7)
• Notable gaps: What the JD asks for that the resume doesn't clearly show
• Best achievements: Top 5–7 bullets ranked for this specific role

Reasoning rules:

• Weight recent roles more heavily than older ones
• Weight ownership language ("built X from scratch", "led a team") over participation language
• Weight quantified claims over unquantified ones, but do not penalize sparse metrics if the context is strong
• If the candidate is switching domains, look for transferable patterns, not perfect keyword matches

Step 4 — Rank Highlights

Call rankHighlights.ts using the scoring rubric (see rubric.md). Score each candidate bullet on:

• Role relevance (25%)
• Business impact (20%)
• Quantitative strength (15%)
• Technical depth (10%)
• Ownership (10%)
• Scale (5%)
• Uniqueness (5%)
• Recency (5%)
• Recruiter appeal (5%)

Select the top 4–6 highlights for slide content. Discard bullets that score < 3.5 total or that cannot be grounded in the source text.

Anti-fabrication rule: Never include a claim that cannot be traced to a specific bullet, project, metric, or phrase in the original resume. If evidence is weak, soften phrasing ("contributed to", "supported", "helped drive") rather than asserting direct ownership.

Step 5 — Choose Positioning Narrative

Call chooseNarrative.ts to generate 3 candidate positioning headlines. Score each on:

• Fit to target role (0–10)
• Clarity for a recruiter reading in 3 seconds (0–10)
• Distinctiveness vs. generic titles (0–10)
• Credibility based on resume evidence (0–10)

Select the highest-scoring angle. Persist in carousel.json.

Mode guidance:

• recruiter: Emphasize clarity, fit signal, seniority, and business impact. Use action + outcome framing.
• hiring-manager: Emphasize technical depth, hard problems, methods, and execution. Show how they work.
• personal-brand: Emphasize distinctiveness, memorable framing, editorial voice. Make it shareable.

Step 6 — Generate 7-Slide Carousel Outline

Call buildOutline.ts. Default structure:

#	Template	Purpose
1	HookSlide	Identity hook — powerful recruiter-facing headline
2	FitSlide	3 concise strengths aligned to the JD
3	ProofSlide	Strongest role-relevant achievement
4	ProofSlide	Second strongest achievement
5	WorkStyleSlide	How they work / capability pattern
6	ValuePropSlide	Value proposition for this role
7	CtaSlide	Open to roles / connect / contact

Step 7 — Write Polished Slide Copy

Call writeSlides.ts. Enforce copy limits:

• Title: 4–9 words
• Subtitle: ≤18 words
• Body: ≤30 words
• Bullets: max 3 per slide
• Metric hero: max 1 per slide

Copy style rules:

• Concise, premium, credible
• Not resume-dump, not buzzword soup
• No vague filler ("results-oriented", "passionate about")
• No invented facts or implied ownership unsupported by resume
• Compress bullets — keep the signal, cut the noise

Step 8 — Render and Export

The pipeline automatically:

1. Renders each slide to HTML using React components + Tailwind
2. Captures PNGs at 1080×1080 (square, LinkedIn-safe)
3. Combines into linkedin_carousel.pdf
4. Writes linkedin_caption.txt with 3 caption variants + hashtags

Output location: output/job-carousel/

Step 9 — Summarize Results

After generation, tell the user:

• Positioning angle chosen and why (score breakdown)
• Top 3 highlights selected with rationale
• Mode used (auto-inferred or user-specified)
• Files generated with paths
• Any gaps noted — what was missing from the resume vs. the JD
• Any softened claims — where evidence was weak and phrasing was deliberately hedged

Error Handling

• If resume is too sparse to score highlights: generate best possible carousel with explicit hedging in body copy; notify user
• If JD is just a title: infer domain priorities from the title, use generic but role-appropriate framing
• If a slide cannot be grounded: substitute a capability-pattern slide or remove it from the output
• If export fails (Playwright/Puppeteer not installed): write HTML files and notify user with install instructions

Output Files

output/job-carousel/
  linkedin_carousel.pdf         # Primary deliverable
  linkedin_carousel.pptx        # Canva-ready (--export canva)
  slides_preview/
    slide_01.png
    slide_02.png
    ...
  figma_export/                 # Figma-ready (--export figma)
    slide_01.svg
    slide_02.svg
    ...
    FIGMA_IMPORT_GUIDE.md
  linkedin_caption.txt          # 3 caption variants
  carousel.json                 # Full structured document
  carousel_manifest.json        # Metadata

Canva Import

1. Run with --export canva
2. Open Canva → Create a design → Import (linkedin_carousel.pptx)
3. Every text box and shape is editable in Canva

Figma Import

1. Run with --export figma
2. Open Figma → drag all figma_export/slide_*.svg files onto the canvas
3. Each slide lands as a 1080×1080 frame with editable text and vector shapes
4. See figma_export/FIGMA_IMPORT_GUIDE.md for detailed instructions