candidate-reviewer

name: candidate-reviewer description: > Evaluate and score job candidates against a job description, producing a detailed self-contained HTML scoring report a hiring manager can open in a browser. Use this skill whenever the user provides resumes (PDFs, text, or files) together with a job description (URL, text, or file) and asks anything about fit, ranking, shortlisting, scoring, comparing applicants, or who to interview — for any role type. Works for Engineering, Support, Product, Design, Engineering Management, Solutions/Sales, Marketing, and HR/Operations roles. Trigger on phrases like: "review these candidates", "score these resumes against the JD", "which candidates match this role", "compare applicants", "rank resumes", "shortlist for interviews", "evaluate against the job posting", "hiring report", "who should we interview", "screen these applicants". Also trigger when a user has a folder of resumes open and asks any question about match or recommendation.

Candidate Review Skill

You produce a professional, structured HTML report that scores and ranks job candidates against a job description. A hiring manager should be able to open it in their browser and make shortlisting decisions immediately.

This skill covers all functions in an IT company: Engineering, Support, Product, Design, Engineering Management, Solutions/Sales, Marketing, and HR/Operations. The same structured workflow applies regardless of the role — dimensions and criteria are always derived from the specific JD.

Step 1 – Parse the Job Description

Accept any format:

• URL → fetch with web_fetch, extract the role requirements from the page
• Text paste or file → read directly

Extract and keep:

• Role title and seniority level
• Role category — see table below; this drives Step 3
• Primary required skills/knowledge areas ("must-haves")
• Secondary or preferred skills ("nice-to-haves")
• Years of experience required
• Key responsibilities (signals what depth matters most)
• Any stated transition pathways or equivalences (e.g., "we welcome Python devs learning Elixir")
• Compensation range if present
• Cultural or mission context (remote-first, startup, NGO, enterprise, etc.)

Step 2 – Classify the Role

Before designing dimensions, identify which of the 8 categories the role falls into. This determines what to look for in resumes and what dimensions to create.

If the role doesn't fit cleanly into one category (e.g., a hybrid "Product + Data" or "Sales + Solutions"), treat it as a hybrid and combine criteria from both categories.

Step 3 – Parse Each Resume

For each candidate, extract the information most relevant to the role category identified in Step 2. Focus your extraction on what the JD cares about.

Universal extractions (all role types):

• Total years of experience + career arc (progression over time)
• Most recent role title, scope, and company context
• Evidence of leadership, mentoring, or people management
• Domain or industry experience (B2B SaaS, fintech, e-commerce, NGO, etc.)
• Communication, writing, or presentation evidence (publications, talks, docs)

Category-specific extractions:

If a skill or area is not mentioned in a resume, note it as unstated (not absent) — score conservatively but don't assume the worst.

Step 4 – Design Scoring Dimensions

Never hardcode dimensions. Derive 7–9 dimensions from what the JD actually signals as important for this specific role. Name them specifically (e.g., "Django/FastAPI Depth" not "Tech Stack", "Product Roadmap Ownership" not "Product Skills") so the report is readable at a glance.

Assign weights that reflect importance — weights must sum to 100%.

General weight guidance by criterion type:

Category-specific dimension ideas (use as inspiration — always derive from the actual JD, never copy blindly):

Engineering: Core Language/Framework, System Architecture, DB & Data Layer, Testing & Code Quality, DevOps/CI-CD, API Design, Seniority, Leadership/Mentorship

Support: Technical Troubleshooting Depth, Customer Communication, Ticket/Process Management, Product Knowledge, SLA/Metric Ownership, Escalation Handling, Seniority

Product: Product Roadmap Ownership, Data-Driven Decision Making, Stakeholder & Exec Communication, Discovery & User Research, Delivery & Execution, Cross-functional Leadership, Domain Fit

Design: Design Process & UX Research, Visual & Interaction Quality, Design Systems, Tool Proficiency (Figma etc.), Cross-functional Collaboration, Communication/Presentation, Seniority

Engineering Management: People Management & Hiring, Delivery Track Record, Technical Credibility, Process & Culture Building, Stakeholder Management, Org Scale/Scope

Solutions/Sales: Technical Product Depth, Customer-Facing Communication, Pre-Sales & Demo Ability, RFP/Proposal Quality, Industry/Domain Fit, AE/PM Collaboration

Marketing: Core Channel Expertise (SEO/Paid/Content/etc.), Campaign Performance & Metrics, Marketing Tools Proficiency, B2B/B2C Domain Fit, Content Quality & Volume, GTM & Launch Experience, Cross-functional Collaboration

HR/Operations: Recruiting or HR Domain Depth, Hiring Metrics & Speed, HRBP or Ops Scope, HR Tools Proficiency, Culture & People Initiatives, Process Design/Improvement, Communication & Influence

Step 5 – Score Each Candidate (1–10 per dimension)

Calibration rules:

• Score based on evidence in the resume, not assumptions. Unstated = 3–4 max.
• If the JD has an explicit transition pathway (e.g., "Ruby/Python devs welcome"), score those candidates 5–7 on the core dimension — not 0.
• Spread your scores. If everyone clusters at 6–7, the report is meaningless. Give 10s to standout candidates and 1s to clear mismatches.
• Calculate weighted score: Σ (score_i × weight_i / 100) → result out of 10.

Step 6 – Write Pros, Cons, and Verdicts

For each candidate write:

• 3–7 pros — concrete, evidence-based strengths specific to this role
• 3–7 cons — concrete gaps or risks specific to this role
• 1-sentence hiring note — the key trade-off or recommendation in plain language
• Verdict — pick one:
  • ✅ Strong Consider — good all-round match, recommend interview
  • ⭐ Top [Skill] Pick — best match on the #1 criterion, even if not #1 overall weighted score
  • 🟡 Consider — worthwhile but has real gaps
  • 🟡 Borderline — marginal fit, proceed with caution
  • 🔴 Weak Match — meets experience bar but wrong skills/domain
  • 🔴 Under-qualified — right direction but too junior or inexperienced
  • 🔴 No Match — wrong function or domain entirely

Use ⭐ when the best match on the single most-critical criterion isn't the #1 ranked candidate overall — this trade-off is valuable for the hiring manager to see.

Step 7 – Generate the HTML Report

Write a single self-contained HTML file — all CSS inline in a <style> tag, no external dependencies, no CDN links. Name it [role-slug]_candidate_scoring.html.

Required sections (in order):

1. Header banner Role title, organization, salary range (if known), count of candidates evaluated, key requirements as styled pill badges.

2. Weight legend All scoring dimensions and their weights shown as colored pills, so readers understand the table before looking at it.

3. Scoring matrix table (full-width, horizontally scrollable if needed)

• Columns: Rank | Candidate | Primary Skills/Background | [one column per dimension with weight in header] | Weighted Score /10
• Score cell colors: ≥8 → bg #C6EFCE text #375623 | 5–7 → bg #FFEB9C text #9C6500 | 1–4 → bg #FFC7CE text #9C0006 | 0 → bg #FF0000 text white
• Weighted score column: bg #1F3864 text white bold, larger font
• Rank badges: gold/silver/bronze filled circles for top 3
• Sticky header, alternating row shading, hover highlight

4. Weighted score bar chart Horizontal bars sorted by score descending. Top-tier bars: #2E5F8A. Mid-tier: #F57C00. Bottom-tier: #E53935. Add a ⭐ annotation on the core-skill specialist if they aren't ranked #1 overall.

5. Pros & Cons cards (responsive 2-column grid)

• Card header: bg #1F3864 text white — rank number, name, primary background, verdict badge
• Card body: two-column — left bg #f0faf0 (pros, green header), right bg #fff5f5 (cons, red header)
• Bullet lists at 12px
• Footer note row: italic, bg #f4f6f9
• Verdict badge colors match their tier (green/amber/red)

6. Scoring criteria guide Table: Dimension | Weight | Scoring guide (1-line scale description per dimension). End with a short paragraph explaining the weight rationale — why is the top criterion weighted highest?

7. Recommendation summary table | Priority tier | Candidate(s) | Justification | Tiers: 🥇 Interview First | 🥈 Strong Interview | 🥉 If Top 3 Unavailable | ⛔ Do Not Proceed

CSS palette

Dark blue:   #1F3864   (headers, nav)
Mid blue:    #2E5F8A   (secondary headers)
Light blue:  #BDD7EE   (accents)
Gold:        #C9A227   (rank badges)
Body bg:     #f4f6f9
Card bg:     white
Alt row:     #f8fafc

Font: Arial throughout. Max-width: 1400px centered. Readable at ≥900px. Print-friendly (cards stack single-column via @media print).

Step 8 – Output

1. Save the HTML file to the user's workspace folder (mounted folder if available, otherwise the outputs folder)
2. Use the present_files tool to surface the file to the user
3. Add a brief text summary (3–5 sentences): how many candidates reviewed, who to interview and why, any standout concern

Edge Cases