1. Overview

This agent screens a resume against a job description and returns a fit score, a breakdown of matched and missing qualifications with evidence from the resume, red flags, and a recommended next action. For candidates who advance, it also generates suggested interview questions based on gaps or areas worth probing.

2. Business value

• Consistent first pass: every resume gets evaluated against the same criteria with the same rigor, which removes the variance that comes with different recruiters having different thresholds.

• Time savings for hiring managers: instead of reading 50 resumes end to end, managers review a scored and annotated shortlist. The evidence citations mean they can verify the agent's reasoning without re-reading the whole resume.

• Better interviews: the suggested questions are tailored to each candidate's specific gaps, so interviewers spend time probing what actually matters rather than running through generic question lists.

3. Inputs

4. Execution steps

1. Parse the job description. Extract the job title, required qualifications, nice-to-have qualifications, minimum years of experience, and any other stated requirements. If the job description doesn't clearly separate required from nice-to-have, use language cues ("must have," "required" vs. "preferred," "bonus," "ideally") to classify them. List what was extracted so the user can verify the interpretation.

2. Parse the resume. Extract work history (company, title, dates, responsibilities), education, skills, certifications, and any other relevant sections.

3. Calculate total relevant experience. Sum the years spent in roles related to the job title. Internships count as half. Freelance/contract work counts fully if the responsibilities are relevant.

4. Match qualifications. For each required and nice-to-have qualification extracted from the job description, search the resume for direct evidence. Mark as:

   • Yes: clear evidence present

   • Partial: related experience but not a direct match

   • No: no evidence found

5. Scan for red flags. Check employment timeline for gaps, short tenures, and inconsistencies.

6. Calculate fit score using the formula in Appendix A.

7. Determine recommended action based on the score ranges in the outputs section.

8. Generate interview questions if the candidate scores 60+.

5. Outputs

Extracted requirements

A list of the qualifications and requirements the agent pulled from the job description, separated into required and nice-to-have. This lets the user confirm the agent interpreted the job description correctly before trusting the rest of the output.

Fit score (0 to 100)

Calculated using the weighted scoring formula in Appendix A.

Matched qualifications

A table showing each qualification from the job description, whether the resume demonstrates it, and the specific evidence.

Missing qualifications

A list of required qualifications where the resume shows no evidence. For each, note whether it might be inferable from related experience (e.g., someone who worked at a company known to use Kubernetes might have exposure even if they don't list it).

Red flags

Things that raise concern, regardless of qualification matching:

• Unexplained gaps longer than 12 months

• Job tenure consistently under 12 months across multiple roles

• Mismatch between job titles and described responsibilities

• Claims that seem inflated relative to role level or company size

• Resume formatting or content issues (e.g., entirely skills-based with no work history)

If no red flags are found, state that explicitly.

Experience summary

A brief paragraph summarizing the candidate's career trajectory, total years of relevant experience, and their apparent areas of depth. This should read like a 30-second verbal summary you'd give a hiring manager.

Suggested interview questions

For candidates scoring 60+, generate 3-5 interview questions targeted at:

• Gaps between required qualifications and resume evidence (to determine if the gap is real or just not listed)

• Areas where evidence is partial or ambiguous

• Red flags that need clarification

• Specific projects or achievements that would be worth digging into

Each question should include a brief note on what you're trying to learn from the answer.

6. Validation checks

• The fit score must be mathematically consistent with the qualification matching results and the weights in Appendix A. Don't estimate.

• Every "Matched? Yes" or "Partial" entry must include a specific citation from the resume, not a vague reference.

• Interview questions must relate to actual gaps or findings from the analysis, not generic questions.

• The extracted requirements must come from the job description text. Don't invent requirements that aren't stated or implied.

7. Edge cases

• Vague job descriptions: if the job description is light on specifics (e.g., "looking for a rockstar engineer"), extract what you can and note which requirements were inferred vs. explicitly stated.

• Resume in a non-standard format (e.g., portfolio link, personal website URL instead of a traditional resume): extract what you can, note the format limitation, and adjust confidence accordingly.

• Overqualified candidates: if the candidate clearly exceeds the requirements (e.g., VP-level applying for a senior IC role), note this as an observation rather than a red flag. It's a conversation starter, not a disqualifier.

• Career changers: if the candidate is transitioning from a different field, evaluate transferable skills and note the career change context. Don't penalize lack of direct industry experience if the skills transfer is strong.

---

Appendix A: Scoring weights

Required qualifications (70% of total score)

Each required qualification is weighted equally. A Yes match earns full points, Partial earns half, No earns zero.

Required Score = (matched_yes + 0.5 × matched_partial) / total_required × 70

Nice-to-have qualifications (15% of total score)

Same formula as required, but the total pool is 15 points.

Nice-to-Have Score = (matched_yes + 0.5 × matched_partial) / total_nice_to_have × 15

If no nice-to-have qualifications were found in the job description, redistribute these 15 points to the required qualifications pool.

Experience and trajectory (15% of total score)

Red flag deductions

The score floors at 0. Deductions can't push it negative.