resume-screening-ai-agent-with-claude-s-skills

内容来源：README.md（说明文档） · 原始地址 · 查看安装指南

原始内容

# Claude Skills Project - Resume Screening

This project demonstrates a complete resume screening workflow using Claude Agent SDK and custom skills. It showcases the **extract → assess → validate → generate** workflow pattern for automated candidate evaluation.
![Resume Screening Workflow](https://github.com/umairalipathan1980/Resume-Screening-AI-Agent-with-Claude-s-Skills/raw/HEAD/images/img.png)

## Overview

The project includes:
- **resume-screener skill**: A systematic workflow for screening and evaluating candidate resumes against job requirements
- **resume_screening_agent.py**: Python script demonstrating how to use the skill programmatically via the Claude Agent SDK
- **Sample data**: Complete job description, evaluation criteria, and five diverse candidate resumes
- Customer **resume-screener** skill in **.claude/skills/resume-scanner/skill.md**.
![Resume Screening Workflow](https://github.com/umairalipathan1980/Resume-Screening-AI-Agent-with-Claude-s-Skills/raw/HEAD/images/img2.png)

## Prerequisites

### System Requirements
- **Python 3.10+** (supports 3.10, 3.11, 3.12, 3.13)
- **Node.js** (version 18 or higher recommended)
- **Claude Code CLI 2.0.0+**

### Installation

1. **Install Claude Code CLI:**
   ```bash
   npm install -g @anthropic-ai/claude-code
   ```

2. **Install Python dependencies:**
   ```bash
   pip install -r requirements.txt
   ```

3. **Set up environment variables:**
   Copy the example file and add your API key:
   ```bash
   cp .env.example .env
   ```
   Edit `.env` and set:
   ```
   CLAUDE_API_KEY=your_api_key_here
   ```

## Resume-Screener Skill

A systematic workflow for screening candidate resumes against job requirements with validation to ensure consistent, objective, and high-quality candidate assessments.

### When to Use This Skill

Use this skill when you need to:
- Screen job applicants and evaluate resumes
- Provide structured candidate assessments
- Generate hiring recommendations with detailed justifications
- Ensure consistent evaluation across multiple candidates

**Trigger phrases:**
- "Screen these resumes"
- "Evaluate these candidates"
- "Review these job applications"
- "Assess candidate qualifications"

### Required Inputs

**Mandatory:**
1. **Job Description** (`inputs/job_description.pdf`) - Complete job posting including requirements, qualifications, responsibilities, and company information
2. **Candidate Resumes** (`resumes/*.pdf`) - Resume files submitted by candidates (PDF or DOCX format)

**Optional (but recommended):**
3. **Evaluation Criteria** (`inputs/evaluation_criteria.pdf`) - Detailed scoring rubric and assessment guidelines
4. **Screening Template** (`inputs/screening_template.pdf`) - Standardized format for presenting evaluation results

> **Note:** If job description or resumes are missing, the skill will request them before proceeding.

### Resume Screening Workflow

The skill processes each resume through a 6-step workflow:

**Step 1: Read and Understand Context**
- Review the job description to understand role requirements
- Review evaluation criteria (if provided) to understand scoring methodology
- Review screening template (if provided) to understand output format
- Read the current candidate's resume thoroughly

**Step 2: Extract and Assess Candidate Information**
- Extract: contact info, education, experience, skills, projects, publications
- Compare each area against job requirements using evaluation criteria
- Note specific evidence from the resume
- Identify strengths and gaps

**Step 3: Generate Initial Evaluation**
- Assign points according to evaluation criteria for each category
- Provide justification based on specific resume content
- Identify 3-5 key strengths and 2-3 concerns
- Provide overall assessment and fit analysis
- Make clear recommendation with suggested next steps

**Step 4: Validate the Evaluation**
- Check completeness, accuracy, consistency, objectivity, clarity, alignment
- Verify each score has supporting evidence from the resume
- Confirm no information was fabricated or assumed
- Ensure scoring is proportional and consistent

**Step 5: Address Validation Failures**
- If validation fails, review feedback and make targeted corrections
- Re-validate after corrections
- Maximum 3 validation rounds

**Step 6: Save Evaluation & Validation Results**
- Format evaluation following screening template (if provided)
- Document validation summary
- Save to `candidate_evaluations.docx`
- All candidates in the SAME file with clear separators

## Usage

### Option 1: Via Claude Code

```
Screen the candidates in the 'resumes/' folder using the 'resume-screener' skill
```

### Option 2: Programmatically (Python)

Run the included Python script:

```bash
python resume_screening_agent.py
```

The script will:
1. Load your API key from `.env`
2. Initialize the Claude Agent with the resume-screener skill
3. Process all resumes in the `resumes/` folder
4. Generate evaluations and save them to `candidate_evaluations.docx`

### Option 3: Via claude.ai or Claude Desktop

1. Zip the entire project folder
2. Go to Settings → Capabilities → Skills
3. Click "Upload skill" and upload the ZIP file
4. In a chat, say: "Screen the candidate resumes using the resume-screener skill"

## How the Agent Works

The Resume Screening Agent is a Python script that leverages the Claude Agent SDK to programmatically invoke the resume-screener skill for evaluating candidate resumes. When executed, the script first loads environment variables from a .env file and sets up authentication by retrieving the Claude API key, which it then maps to the ANTHROPIC_API_KEY environment variable that the SDK expects.

The main function configures the Claude Agent with specific options, including the use of the Claude Sonnet 4.5 model to ensure high-quality evaluation capabilities. The agent is set to load skills and settings from both user and project directories, allowing it to access the resume-screener skill that has been defined in the project structure. The configuration also enables several essential tools including Skill, Read, Write, and Bash, which are necessary for the complete screening workflow to function properly.

Once configured, the agent receives a query prompt that instructs it to screen resumes using the resume-screener skill. The agent then processes this request asynchronously, streaming responses back as it works through each candidate resume. During execution, the agent reads all resumes from the designated folder, evaluates each resume against the job requirements and evaluation criteria, runs validation checks on each evaluation to ensure quality, and generates comprehensive feedback that is ultimately saved to the candidate_evaluations.docx file. Throughout this entire process, each message from the agent is printed to the console in real-time, providing immediate visibility into the screening progress and allowing users to monitor the evaluation workflow as it unfolds.

## Project Structure

```
/skills/
├── .claude/
│   └── skills/
│       └── resume-screener/
│           └── skill.md                    # Resume screener skill definition
├── inputs/                                 # Resume screening inputs
│   ├── job_description.pdf                 # Job posting (required)
│   ├── evaluation_criteria.pdf            # Scoring rubric (optional)
│   └── screening_template.pdf             # Output template (optional)
├── resumes/                                # Candidate resumes
├── candidate_evaluations.docx              # Resume screening output (generated)
├── resume_screening_agent.py               # Python agent script
├── requirements.txt                        # Python dependencies
├── .env.example                            # Example .env file
├── README.md                               # This file
```

## Dependencies

### Python Packages
- **claude-agent-sdk** (>=0.1.6) - Official Anthropic SDK for Claude Agent
- **python-dotenv** (>=1.0.0) - Environment variable management

Install with:
```bash
pip install -r requirements.txt
```

## Getting Started

1. **Clone or download this project**

2. **Install all prerequisites** (Python 3.10+, Node.js, Claude Code CLI)

3. **Install Python dependencies:**
   ```bash
   pip install -r requirements.txt
   ```

4. **Create your `.env` file** with your Claude API key:
   ```bash
   cp .env.example .env
   # Edit .env and add your API key
   ```

5. **Review the sample files** in `inputs/` and `resumes/` directories

6. **Run the screening agent:**
   ```bash
   python resume_screening_agent.py
   ```

7. **Check the output** in `candidate_evaluations.docx`

## Expected Output

The skill generates `candidate_evaluations.docx` containing for each candidate:

1. **Candidate Information** (name, contact, location)
2. **Executive Summary** (2-3 sentences + key strengths/concerns)
3. **Detailed Scoring Table** (all 5 categories with justifications)
4. **Comprehensive Analysis**
   - Technical skills analysis
   - Experience & project analysis
   - Educational background assessment
   - Communication & soft skills evaluation
5. **Key Strengths** (3-5 specific points)
6. **Areas of Concern** (2-3 specific points)
7. **Red Flags / Special Considerations**
8. **Final Recommendation** (Strong Interview / Interview / Maybe / Reject)
9. **Validation Summary** (results from each validation round)

## Customization

### Adapt for Different Roles
- Replace `inputs/job_description.pdf` with your job posting
- Update `inputs/evaluation_criteria.pdf` with your scoring rubric
- Adjust `inputs/screening_template.pdf` with your output format
- Add your candidate resumes to `resumes/` folder

### Extend Functionality
You can modify `resume_screening_agent.py` to:
- Change the AI model (currently `claude-sonnet-4-5`)
- Adjust allowed tools
- Customize the prompt
- Process resumes from different directories
- Export results in different formats

Claude Skills Project - Resume Screening

This project demonstrates a complete resume screening workflow using Claude Agent SDK and custom skills. It showcases the extract → assess → validate → generate workflow pattern for automated candidate evaluation.

Overview

The project includes:

resume-screener skill: A systematic workflow for screening and evaluating candidate resumes against job requirements
resume_screening_agent.py: Python script demonstrating how to use the skill programmatically via the Claude Agent SDK
Sample data: Complete job description, evaluation criteria, and five diverse candidate resumes
Customer resume-screener skill in .claude/skills/resume-scanner/skill.md.

Prerequisites

System Requirements

Python 3.10+ (supports 3.10, 3.11, 3.12, 3.13)
Node.js (version 18 or higher recommended)
Claude Code CLI 2.0.0+

Installation

Install Claude Code CLI:

npm install -g @anthropic-ai/claude-code

Install Python dependencies:
```
pip install -r requirements.txt
```
Set up environment variables: Copy the example file and add your API key:
```
cp .env.example .env
```
Edit .env and set:
```
CLAUDE_API_KEY=your_api_key_here
```

Resume-Screener Skill

A systematic workflow for screening candidate resumes against job requirements with validation to ensure consistent, objective, and high-quality candidate assessments.

When to Use This Skill

Use this skill when you need to:

Screen job applicants and evaluate resumes
Provide structured candidate assessments
Generate hiring recommendations with detailed justifications
Ensure consistent evaluation across multiple candidates

Trigger phrases:

"Screen these resumes"
"Evaluate these candidates"
"Review these job applications"
"Assess candidate qualifications"

Required Inputs

Mandatory:

Job Description (inputs/job_description.pdf) - Complete job posting including requirements, qualifications, responsibilities, and company information
Candidate Resumes (resumes/*.pdf) - Resume files submitted by candidates (PDF or DOCX format)

Optional (but recommended): 3. Evaluation Criteria (inputs/evaluation_criteria.pdf) - Detailed scoring rubric and assessment guidelines 4. Screening Template (inputs/screening_template.pdf) - Standardized format for presenting evaluation results

Note: If job description or resumes are missing, the skill will request them before proceeding.

Resume Screening Workflow

The skill processes each resume through a 6-step workflow:

Step 1: Read and Understand Context

Review the job description to understand role requirements
Review evaluation criteria (if provided) to understand scoring methodology
Review screening template (if provided) to understand output format
Read the current candidate's resume thoroughly

Step 2: Extract and Assess Candidate Information

Extract: contact info, education, experience, skills, projects, publications
Compare each area against job requirements using evaluation criteria
Note specific evidence from the resume
Identify strengths and gaps

Step 3: Generate Initial Evaluation

Assign points according to evaluation criteria for each category
Provide justification based on specific resume content
Identify 3-5 key strengths and 2-3 concerns
Provide overall assessment and fit analysis
Make clear recommendation with suggested next steps

Step 4: Validate the Evaluation

Check completeness, accuracy, consistency, objectivity, clarity, alignment
Verify each score has supporting evidence from the resume
Confirm no information was fabricated or assumed
Ensure scoring is proportional and consistent

Step 5: Address Validation Failures

If validation fails, review feedback and make targeted corrections
Re-validate after corrections
Maximum 3 validation rounds

Step 6: Save Evaluation & Validation Results

Format evaluation following screening template (if provided)
Document validation summary
Save to candidate_evaluations.docx
All candidates in the SAME file with clear separators

Usage

Option 1: Via Claude Code

Screen the candidates in the 'resumes/' folder using the 'resume-screener' skill

Option 2: Programmatically (Python)

Run the included Python script:

python resume_screening_agent.py

The script will:

Load your API key from .env
Initialize the Claude Agent with the resume-screener skill
Process all resumes in the resumes/ folder
Generate evaluations and save them to candidate_evaluations.docx

Option 3: Via claude.ai or Claude Desktop

Zip the entire project folder
Go to Settings → Capabilities → Skills
Click "Upload skill" and upload the ZIP file
In a chat, say: "Screen the candidate resumes using the resume-screener skill"

How the Agent Works

The Resume Screening Agent is a Python script that leverages the Claude Agent SDK to programmatically invoke the resume-screener skill for evaluating candidate resumes. When executed, the script first loads environment variables from a .env file and sets up authentication by retrieving the Claude API key, which it then maps to the ANTHROPIC_API_KEY environment variable that the SDK expects.

The main function configures the Claude Agent with specific options, including the use of the Claude Sonnet 4.5 model to ensure high-quality evaluation capabilities. The agent is set to load skills and settings from both user and project directories, allowing it to access the resume-screener skill that has been defined in the project structure. The configuration also enables several essential tools including Skill, Read, Write, and Bash, which are necessary for the complete screening workflow to function properly.

Once configured, the agent receives a query prompt that instructs it to screen resumes using the resume-screener skill. The agent then processes this request asynchronously, streaming responses back as it works through each candidate resume. During execution, the agent reads all resumes from the designated folder, evaluates each resume against the job requirements and evaluation criteria, runs validation checks on each evaluation to ensure quality, and generates comprehensive feedback that is ultimately saved to the candidate_evaluations.docx file. Throughout this entire process, each message from the agent is printed to the console in real-time, providing immediate visibility into the screening progress and allowing users to monitor the evaluation workflow as it unfolds.

Project Structure

/skills/
├── .claude/
│   └── skills/
│       └── resume-screener/
│           └── skill.md                    # Resume screener skill definition
├── inputs/                                 # Resume screening inputs
│   ├── job_description.pdf                 # Job posting (required)
│   ├── evaluation_criteria.pdf            # Scoring rubric (optional)
│   └── screening_template.pdf             # Output template (optional)
├── resumes/                                # Candidate resumes
├── candidate_evaluations.docx              # Resume screening output (generated)
├── resume_screening_agent.py               # Python agent script
├── requirements.txt                        # Python dependencies
├── .env.example                            # Example .env file
├── README.md                               # This file

Dependencies

Python Packages

claude-agent-sdk (>=0.1.6) - Official Anthropic SDK for Claude Agent
python-dotenv (>=1.0.0) - Environment variable management

Install with:

pip install -r requirements.txt

Getting Started

Clone or download this project
Install all prerequisites (Python 3.10+, Node.js, Claude Code CLI)
Install Python dependencies:
```
pip install -r requirements.txt
```

Create your .env file with your Claude API key:

cp .env.example .env
# Edit .env and add your API key

Review the sample files in inputs/ and resumes/ directories
Run the screening agent:
```
python resume_screening_agent.py
```
Check the output in candidate_evaluations.docx

Expected Output

The skill generates candidate_evaluations.docx containing for each candidate:

Candidate Information (name, contact, location)
Executive Summary (2-3 sentences + key strengths/concerns)
Detailed Scoring Table (all 5 categories with justifications)
Comprehensive Analysis
- Technical skills analysis
- Experience & project analysis
- Educational background assessment
- Communication & soft skills evaluation
Key Strengths (3-5 specific points)
Areas of Concern (2-3 specific points)
Red Flags / Special Considerations
Final Recommendation (Strong Interview / Interview / Maybe / Reject)
Validation Summary (results from each validation round)

Customization

Adapt for Different Roles

Replace inputs/job_description.pdf with your job posting
Update inputs/evaluation_criteria.pdf with your scoring rubric
Adjust inputs/screening_template.pdf with your output format
Add your candidate resumes to resumes/ folder

Extend Functionality

You can modify resume_screening_agent.py to:

Change the AI model (currently claude-sonnet-4-5)
Adjust allowed tools
Customize the prompt
Process resumes from different directories
Export results in different formats