@arkone_ai/cv-batch-screen 1.0.0

package/README.md

@@ -0,0 +1,29 @@
+# @arkone_ai/cv-batch-screen
+
+A [Claude Code](https://claude.ai/code) skill that screens multiple candidate CVs against a job description, ranks them, and produces a comparison report.
+
+## Install
+
+```bash
+npm install -g @arkone_ai/cv-batch-screen
+```
+
+## Usage
+
+In any Claude Code session:
+
+```
+/cv-batch-screen
+```
+
+## What it does
+
+- Accepts a folder of CVs (PDF, DOCX, images) and a job description
+- Scores each candidate across 5 categories with the same rubric as `/cv-screen`
+- Ranks all candidates and categorizes as Advance, Hold, or Pass
+- Produces a comparison table with per-candidate profiles for top candidates
+- Supports export as Markdown, CSV, JSON, or branded PDF
+
+## License
+
+MIT

package/install/install.js

@@ -0,0 +1,38 @@
+#!/usr/bin/env node
+
+import { copyFileSync, mkdirSync, existsSync } from 'fs';
+import { join } from 'path';
+import { homedir } from 'os';
+
+const root = join(import.meta.dirname, '..');
+const home = homedir();
+
+const agents = [
+  { name: 'Claude Code', dir: join(home, '.claude', 'skills') },
+  { name: 'Codex',       dir: join(home, '.codex',  'skills') },
+  { name: 'Kiro',        dir: join(home, '.kiro',   'skills') },
+];
+
+const files = [
+  { src: join(root, 'skill.md'), dest: 'cv-batch-screen.md' },
+];
+
+let installed = 0;
+for (const agent of agents) {
+  if (!existsSync(agent.dir)) continue;
+  for (const file of files) {
+    copyFileSync(file.src, join(agent.dir, file.dest));
+  }
+  console.log(`✓ ${agent.name}: ${agent.dir}`);
+  installed++;
+}
+
+if (installed === 0) {
+  mkdirSync(agents[0].dir, { recursive: true });
+  for (const file of files) {
+    copyFileSync(file.src, join(agents[0].dir, file.dest));
+  }
+  console.log(`✓ Claude Code: ${agents[0].dir}`);
+}
+
+console.log('\nUse /cv-batch-screen in any supported AI session.');

package/package.json

@@ -0,0 +1,27 @@
+{
+  "name": "@arkone_ai/cv-batch-screen",
+  "version": "1.0.0",
+  "description": "Claude Code skill: batch screen multiple CVs against a job description, rank and compare candidates",
+  "type": "module",
+  "scripts": {
+    "postinstall": "node install/install.js"
+  },
+  "bin": {
+    "cv-batch-screen": "install/install.js"
+  },
+  "files": [
+    "skill.md",
+    "install/"
+  ],
+  "keywords": [
+    "claude",
+    "claude-code",
+    "arkone",
+    "hr-ops",
+    "cv-batch-screen"
+  ],
+  "license": "MIT",
+  "engines": {
+    "node": ">=18"
+  }
+}

package/skill.md

@@ -0,0 +1,177 @@
+---
+name: cv-batch-screen
+description: "Screen multiple candidate CVs/resumes against a job description, rank them, and produce a comparison report. Use this skill when: 'screen these CVs', 'rank candidates', 'compare resumes', 'batch screen', 'review all applicants', 'shortlist candidates'. Also trigger for: 'candidate ranking', 'bulk CV review', 'applicant comparison'."
+license: MIT
+compatibility: Designed for Claude Code
+---
+
+# Skill: cv-batch-screen
+
+Screen multiple candidate CVs against a job description, score each one, rank them, and produce a comparison table with recommendations.
+
+## When to use
+
+- User has multiple CVs and wants them ranked against a JD
+- User wants to shortlist candidates from a pool
+- User wants a side-by-side comparison of candidates
+
+## Step 1: Gather inputs
+
+### Required
+
+| Input | Accepts |
+|---|---|
+| CVs / Resumes | Folder path, glob pattern (e.g., `./candidates/*.pdf`), or list of file paths |
+| Job Description | File path (PDF, DOCX, TXT) or pasted text |
+
+### Optional
+
+| Input | Default | Purpose |
+|---|---|---|
+| Must-have skills | Extracted from JD | Override or supplement |
+| Shortlist size | Top 5 or all | How many candidates to advance |
+| Priority weights | Equal | Custom category weights |
+| Screening strictness | Standard | `strict` / `standard` / `lenient` |
+| Output format | Markdown | `md`, `csv`, `json`, `pdf` |
+| Auto-reject threshold | 40 | Score below which candidates are auto-rejected |
+
+### Collecting CVs
+
+1. If given a folder path: glob for `*.pdf`, `*.docx`, `*.doc`, `*.txt`, `*.png`, `*.jpg`
+2. If given a glob pattern: expand it
+3. If given individual files: use as-is
+4. Report how many CVs were found before processing
+
+```
+Found 12 CVs in ./candidates/
+Processing...
+```
+
+## Step 2: Parse JD (once)
+
+Use the same JD parsing logic as the `cv-screen` skill:
+- Extract role title, level, must-have skills, nice-to-haves, experience requirements, education requirements, key responsibilities
+
+This is done once and reused for all candidates.
+
+## Step 3: Screen each candidate
+
+For each CV, run the same analysis as the `cv-screen` skill:
+
+1. Parse the CV — extract name, experience, skills, education, work history
+2. Score against JD — skills match, experience, education, career trajectory, role alignment
+3. Detect red flags
+4. Calculate overall score
+
+### Progress reporting
+
+Show progress as candidates are processed:
+
+```
+[1/12] Jane Smith ........... 87/100 ✅
+[2/12] John Doe ............. 72/100 ⚠️
+[3/12] Alex Chen ............ 91/100 ✅
+[4/12] Pat Johnson .......... 35/100 ❌
+...
+```
+
+## Step 4: Rank and categorize
+
+### Ranking
+
+Sort all candidates by overall score (descending). Break ties by:
+1. Skills match score (higher is better)
+2. Experience score (higher is better)
+3. Fewer red flags
+
+### Categories
+
+| Category | Criteria | Action |
+|---|---|---|
+| **Advance** ✅ | Score ≥ 75 | Move to interview |
+| **Hold** ⚠️ | Score 40-74 | Review if pipeline is thin |
+| **Pass** ❌ | Score < 40 (or below auto-reject threshold) | Do not advance |
+
+## Step 5: Generate output
+
+### Summary table
+
+```
+📋 Batch CV Screening Report
+━━━━━━━━━━━━━━━━━━━━━━━━━━━━
+Role:        Senior Software Engineer
+Candidates:  12 screened
+Date:        2026-03-23
+
+── Ranking ──
+
+| Rank | Candidate      | Score | Skills | Exp  | Edu  | Trajectory | Alignment | Flags | Action  |
+|------|----------------|-------|--------|------|------|------------|-----------|-------|---------|
+| 1    | Alex Chen      | 91    | 28/30  | 23/25| 15/15| 13/15      | 12/15     | 0     | ✅ Adv  |
+| 2    | Jane Smith     | 87    | 27/30  | 22/25| 13/15| 14/15      | 11/15     | 0     | ✅ Adv  |
+| 3    | Sam Rivera     | 82    | 25/30  | 21/25| 12/15| 12/15      | 12/15     | 1     | ✅ Adv  |
+| 4    | John Doe       | 72    | 22/30  | 18/25| 12/15| 10/15      | 10/15     | 1     | ⚠️ Hold |
+| ...  | ...            | ...   | ...    | ...  | ...  | ...        | ...       | ...   | ...     |
+| 12   | Pat Johnson    | 35    | 10/30  | 8/25 | 7/15 | 5/15       | 5/15      | 3     | ❌ Pass |
+
+── Summary ──
+✅ Advance:  5 candidates
+⚠️ Hold:     4 candidates
+❌ Pass:      3 candidates
+```
+
+### Per-candidate details
+
+For each candidate in the Advance category, include a brief profile:
+
+```
+── #1 Alex Chen (91/100) ──
+Current: Staff Engineer at Stripe (4 years)
+Strengths: Full-stack match, K8s + AWS certified, led team of 8
+Gaps: No Go experience (nice-to-have)
+Interview focus: System design depth, management transition interest
+Red flags: None
+```
+
+For Hold candidates, include a one-liner explaining what's missing:
+```
+── #4 John Doe (72/100) — Hold ──
+Strong backend skills but no cloud infrastructure experience (must-have). Consider if role can flex.
+```
+
+For Pass candidates, one line each:
+```
+── #12 Pat Johnson (35/100) — Pass ──
+Career focused on data science, minimal overlap with backend engineering requirements.
+```
+
+### Save report
+
+Default: `./cv-batch-screen-<role-slug>-<date>.md`
+
+Also support:
+- `--format csv` → save ranking table as CSV (for spreadsheet import)
+- `--format json` → structured JSON (for ATS integration)
+- `--format pdf` → branded PDF report (uses brand config like offer-letter skill)
+
+## Step 6: Optional follow-up actions
+
+After presenting results, offer these next steps:
+
+| Action | What it does |
+|---|---|
+| "Screen more" | Add more CVs to the existing batch and re-rank |
+| "Adjust weights" | Re-score with different category weights |
+| "Generate interview kits" | Create interview scorecards for Advance candidates |
+| "Export for ATS" | Export as JSON/CSV for ATS import |
+| "Email rejections" | Draft and send polite rejection emails to Pass candidates (requires email config) |
+
+## Error handling
+
+| Issue | What to do |
+|---|---|
+| No CVs found in path | Ask user to verify the path or provide files |
+| Mixed file types | Process what's readable, report which files failed |
+| Duplicate candidates | Detect by name similarity, keep the most recent CV, warn user |
+| Very large batch (50+) | Warn about processing time, suggest screening in chunks |
+| CV in foreign language | Note the language, attempt analysis, flag confidence as lower |