moai-recruiter | Korean hiring workflow. AI only scores; humans decide.
⚠️ Mandatory Safety Principles (HARD)
- Automatic rejection absolutely prohibited: This skill only provides scores and rationale and does not output a "reject/pass" label
- Human review required: Automatically inserts the header "This evaluation is for decision support. The final decision is made by a human reviewer" at the top of all outputs
- Protected-category masking: Automatically hides photo, date of birth, place of origin, family relationships, physical information, religion, and political orientation
- 4-year log retention: Automatically records input hash, evaluation time, and job description ID used → the user stores them separately
- Biased keyword check: Warns when discriminatory expressions such as "young talent", "male employees only", or "whether planning to marry" are entered
Violation case (reference): iTutorGroup $365K settlement — ADEA violation due to age-based automatic rejection.
Overview
This is a resume evaluation skill designed to simultaneously comply with Korean and global regulations, including the Fair Hiring Procedure Act (2014, applicable to workplaces with 30 or more employees), the NCS classification system, EEOC Title VII, and NYC Local Law 144. The AI provides suitability scores and cited rationale for each NCS competency unit, and the hiring manager uses these as decision-making materials.
Trigger keywords
resume review applicant evaluation document screening resume analysis NCS job fit hiring document review resume score
Workflow
Step 1: Input collection
- Job description (required): Recommended based on the NCS classification system (if absent,
moai-recruiter:hr-employmentguide the user to first create a job description with) - Resume (required): Text/PDF
- Cover letter (optional)
- Evaluation weights (optional): Weight per competency unit in the range 0-3
Step 2: Automatic protected-category masking (required, cannot be disabled by the user)
The following information is automatically hidden before evaluation:
| Category | Masking method |
|---|---|
| Photo | Ignore (does not parse images) |
| Date of birth / age | Graduation year → show education only |
| Place of origin / family origin | Show only up to metropolitan city/province |
| Family relationships | Fully masked |
| Physical information (height/weight) | Fully masked |
| Religion / political orientation | Fully masked |
| Marital status / children | Full masking |
| SNS with photo and name exposed | Display URL only, ignore own photo |
Step 3: Scoring by NCS competency unit
After extracting the competency units (Unit of Competence) for the relevant job from the NCS classification system, assign 0-3 points to each:
| Score | Meaning |
|---|---|
| 0 | No relevant experience or learning evidence |
| 1 | Only indirect experience or introductory learning confirmed |
| 2 | Direct experience or intermediate learning confirmed |
| 3 | Sufficient direct experience + quantitative results confirmed |
Each score is resume citation(direct text excerpt) presented together. Scores without citations are invalid.
Step 4: Summary of strengths, weaknesses, and items requiring confirmation
- 3 strengths: extracted from competency units with a score of 2-3
- 3 weaknesses: competency units with a score of 0-1 + core job mismatch areas
- 3 items requiring additional confirmation: candidate questions to confirm in the interview
Step 5: Decision-support information
Output the following (labels and automatic recommendations prohibited):
- NCS competency unit match rate (%)
- Job core competency fit (qualitative)
- 5 candidate priority questions for follow-up interviews
Prohibited: Never output labels such as "Strong Match", "Recommend: Interview", "Not Suitable", "Recommend Rejection".
Step 6: Automatic audit trail recording
Add audit information at the end of the output:
## 감사 정보 (4년 보관 권장)
- 평가 시각: 2026-05-01T12:00:00+09:00
- 입력 이력서 SHA256: <hash>
- 사용 직무기술서 ID: JD-2026-001
- 마스킹 처리 항목: 사진, 생년월일, 가족관계, 결혼 여부
- AI 모델: claude-opus-4-7
- 사용자: <orchestrator user>
Output format
> ⚠️ 본 평가는 의사결정 보조용입니다. 최종 결정은 인간 검토자가 수행해야 합니다.
> 한국 채용절차법 및 관련 규제(EEOC Title VII, NYC Local Law 144 등)에 따라 자동 거절·자동 합격은 금지됩니다.
## NCS 능력단위별 적합도
| 능력단위 | 점수(0-3) | 인용 근거 |
|---|---|---|
| 시스템분석 | 3 | "AWS 마이그레이션 프로젝트 PM 24개월" |
| ... | ... | ... |
매칭률: 72% (능력단위 합계 / 최대 가능 점수)
## 강점 (3)
1. ...
2. ...
3. ...
## 약점 (3)
1. ...
...
## 면접 확인 필요 사항 (3)
1. ...
...
## 추천 면접 질문 후보 (5)
1. ...
...
## 감사 정보 (4년 보관 권장)
- 평가 시각: ...
- 입력 이력서 SHA256: ...
- ...
Usage examples
Example 1 — NCS job description + 1 resume
사용자: "백엔드 시니어 채용. NCS 직무기술서 첨부, 지원자 이력서 PDF 첨부."
→ hr-resume-screener 가 능력단위별 0-3 점수 + 인용 + 강점/약점/면접 질문
→ 자동 거절·자동 합격 라벨 없음, 감사 정보 자동 기록
Example 2 — Protected-class masking demonstration
사용자: "이 이력서 평가해줘. (사진·주민번호 앞자리 포함)"
→ hr-resume-screener 가 사진·생년월일·출신지 자동 가린 후 평가
→ 출력 상단에 "마스킹 처리 항목: ..." 명시
Precautions
- Disclaimer of legal liability: This skill is a decision-support tool, and all legal responsibility for hiring decisions rests with the user (employer)
- Awareness of iTutorGroup·Mobley v. Workday case: AI tool developers can also be held liable under Title VII. This skill is designed not to output automated decisions
- Application to workplaces with 30 or more employees: The Korean Fair Hiring Procedure Act applies to workplaces with 30 or more full-time employees. Companies with fewer than 30 employees are also recommended to follow the same principles
- When operating a talent pool: Collect only minimal identifying information (name, contact details). Storing career history, certifications, and self-introductions in a talent pool is prohibited (risk of evaluation that circumvents the Fair Hiring Procedure Act)
- Bias audit: If the user is subject to NYC Local Law 144, etc., a separate annual bias audit must be conducted (Greenhouse is a case of Warden AI monthly audits)
- Prohibition of time estimates: Hiring schedules should use priority labels (e.g., "P0: recommend interview within 1 week")
Related skills
Before (pre-evaluation):
moai-recruiter:hr-employment— Write the NCS job description (JD) first. Define the job requirements that will serve as the evaluation criteria first.
After (post-evaluation):
moai-recruiter:hr-employment— Based on the scores and items requiring confirmation output by this skill, it constructs interview questions and interview design (including the weighting of job competencies and cultural fit).- Cultural fit and in-depth cover letter analysis are handled in this skill's cover letter evaluation stage (steps 2-4) without a separate skill, and cultural fit verification in the interview stage is
moai-recruiter:hr-employmentleads to the interview design of.
Alternative:
- External ATS (Greenhouse, Lever, Workday) — This skill is an aid, not a replacement for ATS
Source
- NCS National Competency Standards — Competency unit classification system
- Yulchon LLC — Employment Procedure Act — Applies to 30 or more employees, document destruction obligation, talent pool operation
- Hwawoo Law Firm — AI recruitment regulation trends
- Korea Labor Institute — AI HR and labor report
- Greenhouse AI/ML Security & Privacy — ISO 42001:2023, Warden AI monthly bias audit
- Workforce Bulletin — Mobley v. Workday — Title VII liability for AI tool developers
- HR Defense — AI Hiring Compliance 2026
- DISA — AI in HR 2026 Risks