Resume Screening Template#
The Resume Screening Template is designed for building an agent that analyzes resumes, evaluates candidate fit against job requirements, and determines whether mandatory qualifications are met. Based on registered resume documents, it systematically analyzes experience, tech stack, qualifications, and more to streamline the recruiter’s initial screening process.High-volume resume processing: Register dozens to hundreds of resumes and quickly identify candidates who meet the criteria through queries
Criteria-based evaluation: Define required qualifications, preferred qualifications, and other criteria for each role in the prompt to ensure consistent evaluation
Evidence-based judgment: Present evaluation rationale together with specific excerpts from the original resume
Flexible criteria updates: Even when the role changes, new criteria can be applied immediately by updating only the prompt
The Resume Screening Template is used after registering resumes as Knowledge. After creating the agent, upload the resume files on the Knowledge page.
Preparation#
After creating the agent, prepare the following two items before performing actual screening.1. Register Resumes#
Upload the resumes to be screened on the Knowledge page.Supported formats: Common document formats such as PDF and DOCX
If there are multiple resumes, you can upload them in bulk at once.
You can use folders to organize and manage resumes by role or job posting.
If you are concerned about sensitive information in resumes, such as resident registration numbers or contact details,
you can register review patterns on the Training Data Quality Management page to detect sensitive information in advance.
2. Define Evaluation Criteria#
On the Workflow page, update the system prompt of the LLM node to define the evaluation criteria.
Please specify required qualifications, preferred qualifications, evaluation categories, output format, and so on for the target role.You are an IT recruiting expert. Please evaluate the candidate’s suitability based on the following criteria.Required qualifications: 3+ years of Java/Spring experience, degree in Computer Science or related field
Preferred qualifications: AWS experience, experience handling large-scale traffic
Evaluation categories: technical fit, years of experience, project experience
Output format: fit / not fit decision, category scores (out of 5), rationale for the decision
Workflow Structure#
The default workflow structure of the Resume Screening Template is as follows.Step 1. Candidate Profile Summary (LLM)#
Extracts only the key candidate information from the resume and generates a structured summary.
This summary is used as reference material in later evaluation steps.Name, position, total experience, key skills
Most recent experience (company name, employment period)
Name: Hong Gil-dong
Position: Frontend Developer
Total Experience: 5 years 2 months
Key Skills: React, Next.js, TypeScript
Most Recent Experience: Microsoft (2020.12 - 2023.03, 2 years 4 months)
Step 1-B. Risk / Caution Review (LLM) — Runs in Parallel with Step 1#
Runs at the same time as Step 1 and identifies risk factors in the resume from the recruiter’s perspective.Mismatch in years of experience (total experience vs. hands-on experience vs. relevant experience)
Frequent job changes (repeated short tenures, career gaps)
Conflicting information in the resume (contradictions / unclear sections)
Step 2. Category-by-Category Evaluation (LLM)#
Compares the JD and the resume and evaluates four categories individually.
It also references the candidate profile summary generated in Step 1.| Evaluation Category | Evaluation Standard | Result |
|---|
| Requirements | Whether the JD’s mandatory qualifications are met | Insufficient / Needs Review / Meets |
| Preferred Qualifications | Whether the JD’s preferred qualifications match | Insufficient / Needs Review / Meets |
| Job Fit | Relevance of experience, role, achievements, and skills to the JD | Insufficient / Needs Review / Meets |
| Filter Check | Whether predefined rejection / hold conditions apply | Pass / Fail |
The rationale for each category is output in the following structure:Condition: Requirement stated in the JD
Resume Item: Actual content stated in the resume (including numbers)
Judgment: Logical explanation of why the decision was made
Default Filter Check Conditions:Experience requirement mismatch (below minimum / above maximum)
Job domain mismatch (experience completely unrelated to the JD)
Missing core competencies (no required qualifications / skills at all)
Organizational fit structure mismatch (collaboration environment mismatch)
Scope imbalance (experience limited to only a specific sub-area)
Step 3. Pass / Fail Decision (LLM)#
Combines the evaluation results from Step 2 to make the final decision.| Condition | Final Result |
|---|
| Filter check failed | Reject |
| Filter check passed + Requirements or Job Fit = "Insufficient" | Reject |
| Mostly "Needs Review" overall, with mixed strengths and weaknesses | Needs Review |
| Core categories = "Meets" + Preferred Qualifications = at least "Needs Review" | Proceed to Interview |
| All core categories = "Meets" | Proceed to Interview |
At this stage, a summary report including the rationale for each category is also generated.Step 4. Conditional Branching (If/Else)#
The workflow branches based on the decision result from Step 3.Proceed to Interview or Needs Review → Move to Step 5 (Interview Question Generation)
Reject → Skip Step 5 and move directly to Result Combination
Step 5. Interview Question Generation (LLM) — Runs Only When Passed#
Based on the JD and the resume analysis results, generates 5 customized interview questions.Question Generation Criteria:Questions that can verify risks identified in the caution review
In-depth questions about the experience / projects described in the resume
Technical questions related to the core competencies in the JD
Situational questions to assess problem-solving ability
Result Combination (Response Node)#
Combines all analysis results into a single response.Resume Screening Result: [Proceed to Interview / Needs Review / Reject]Category-by-category evaluation summary report
Candidate profile summary
Recommended Interview Questions ← Included only when passed(5 interview questions + question intent)
Even if the job posting changes, there is no need to upload the resumes again.
You can immediately re-evaluate them against the new role criteria by updating only the evaluation criteria in the system prompt of the LLM node.
If you organize resumes into folders by role and set the knowledge search scope of the search node to a specific folder,
you can perform accurate screening for applicants for that role only.
