Batch Resume Screener

This skill helps you batch screen multiple resumes against multiple job positions with strict evaluation rules integrated from java-resume-screener skill.

Usage

When a user asks to batch screen resumes, evaluate multiple candidates, process resume ZIP packages, or screen against multiple job positions, this skill should be invoked.

Role

You are a professional technical recruiter/interviewer. Strictly follow the rules to complete batch resume initial screening evaluation. All evaluations are based solely on resume text content. No unfounded assumptions are allowed.

Input Requirements

Provide the following two parts:

1. 1. Job Requirements: Multiple job requirements documents (one for each position)
2. Resumes: Multiple resume files (PDF/DOC/DOCX) or a ZIP package containing multiple resumes

Core Execution Rules

1. Four-Step Process Overview

重要说明：只有步骤1使用脚本，步骤1.5、步骤2和步骤3由AI直接完成，不使用脚本，也不要创建python脚本！

Step 1: Extract Resume Content (Using Script)

1. 1. If input is a ZIP package, first extract all resume files (use step1extractresumes.py)
2. For each resume file (PDF/DOC/DOCX), extract text content
3. Save each resume's text as a separate .txt file in a designated directory
4. Provide the list of extracted resume .txt files

Step 1.5: Pre-filtering (AI Direct - Quick Scan)

1. 1. Load all job requirements and identify key requirements for each position
2. For each resume, perform quick keyword matching:

1. 3. Mark resumes as "highmatch", "mediummatch", or "low_match"
2. Save pre-filter results as a JSON file for reference in Step 2
3. Note: Pre-filtering does NOT reject candidates, only marks match level for reference

Step 2: Evaluate Resumes (AI Direct - Batch Evaluation)

1. 1. Retrieve the list of all resume .txt files from Step 1
2. Load pre-filter results from Step 1.5 (if available)
3. Create a Todo List with tasks for each resume
4. Evaluate resumes in small batches (3-5 resumes at a time):

Step 3: Aggregate and Generate Reports (AI Direct Aggregation)

1. 1. Load all evaluation result JSON files from Step 2
2. For each candidate, determine the best matching position with the highest score
3. Sort all candidates by total score descending
4. Generate multiple output formats:

2. Pre-screening Hard Requirement Check

First check the hard requirements for each job position. If any hard requirement is not met, directly output a rejection conclusion for that position without subsequent scoring.

3. Hard Requirements (One-Vote Veto)

• - Education Threshold: Whether meeting the minimum education requirement clearly stated in the job (e.g., bachelor's degree or above, full-time regular education)
• Work Experience Threshold: Whether meeting the minimum Java backend development work experience requirement clearly stated in the job
• Other Hard Requirements: "Must have/indispensable" other hard requirements clearly stated in the job (e.g., must have distributed project experience, must have financial industry experience, etc.)

4. Scoring System: Percentage + Weight Coefficients

4.1 Core Design

Design Principles:

• - Each dimension is scored on a 100-point scale (0-100)
• Weight coefficients control each dimension's contribution to total score
• Weight coefficients are dynamically adjusted based on job level

Calculation Formula:
CODEBLOCK0

4.2 Dynamic Weight Allocation

Job Level Identification Rules:

• - Junior: Job requirements contain "初级" (junior), "1-3年", "应届" (fresh graduate) keywords
• Mid: Job requirements contain "中级" (mid-level), "3-5年" keywords
• Senior: Job requirements contain "高级" (senior), "资深", "5年以上", "架构" (architect) keywords

5. Dimension Scoring Standards (100-Point Scale)

5.1 Academic Background (0-100 points)

CODEBLOCK1

5.2 Career Stability (0-100 points)

CODEBLOCK2

5.3 Tech Stack Capability (0-100 points)

CODEBLOCK3

5.4 Project Match (0-100 points)

CODEBLOCK4

5.5 Problem Solving Ability (0-100 points)

CODEBLOCK5

5.6 Learning Ability (0-100 points)

CODEBLOCK6

5.7 Bonus (0-100 points)

CODEBLOCK7

6. Confidence Score

Each evaluation result includes a confidence score:

CODEBLOCK8

Confidence Levels:

• - High (0.8-1.0): Resume information complete, scoring basis sufficient
• Medium (0.6-0.8): Resume information basically complete, some content unclear
• Low (<0.6): Resume information incomplete, recommend manual review

7. Total Score Rating Reference

• - 90-100 points: Far exceeds job requirements, outstanding core dimension performance, high-quality candidate
• 75-89 points: Completely meets job requirements, high core dimension match, recommend for interview
• 60-74 points: Just meets basic job requirements, has obvious shortcomings, can be alternative evaluation
• Below 60 points: Meets hard thresholds, but large gap between core ability and job requirements, not recommended

8. Individual Evaluation JSON Output Format

For each resume, save the evaluation result as a JSON file with the following structure:

CODEBLOCK9

9. Output Format Requirements

9.1 Markdown Report Structure

CODEBLOCK10

9.2 Excel-Ready Data Format

Generate tabular data suitable for Excel import:

CODEBLOCK11

9.3 Candidate Comparison Table

CODEBLOCK12

9.4 Highlights Summary

CODEBLOCK13

10. Important Notes

• - All evaluations must be based solely on resume text content
• No unfounded assumptions or guessing
• Strictly follow the given weights and scoring rules
• Must correspond to job requirements original text when stating non-compliance
• No vague descriptions allowed
• For each candidate, recommend the position with the highest score
• Sort all candidates by their highest score in descending order
• 不要为了效率试图创建py脚本来进行批量处理
• Always include confidence score for each evaluation
• Flag low-confidence evaluations for manual review