Resume Coach: Interactive Resume Improvement Process
Transform resumes from generic AI-polished documents into compelling, personalized stories that pass recruiter screening. This skill implements a coaching methodology that helps users discover and articulate their unique value proposition.
Overview
Most AI-assisted resume improvements produce similar results because they lack context. This skill takes a different approach: instead of simply rewriting, it guides users through a structured discovery process that uncovers hidden strengths and creates genuine differentiation.
Core Philosophy: The goal is not to write the resume FOR the user, but to help them rediscover and reframe their own experiences.
Required Inputs
Before starting the process, collect:
- Resume: User's current resume (file path or pasted text)
- Job Posting: Target position description (URL, file, or pasted text)
If either is missing, request it before proceeding.
The Coaching Process
Phase 1: Recruiter Perspective Analysis
Analyze the resume from a hiring manager's viewpoint. Identify 3-5 questions a recruiter would ask when reviewing this resume.
Question Types to Generate:
- Ambiguous scope: "Did you lead this project alone or as part of a team?"
- Missing metrics: "What was the actual impact of this initiative?"
- Unclear responsibilities: "Were you responsible for A, B, or both?"
- Skill depth: "How extensively did you use this technology?"
- Gap filling: "What happened between these two positions?"
Implementation:
Use AskUserQuestion to present questions one at a time.
Collect answers to enrich the resume context.
Store responses for use in final resume.
Phase 2: Ideal Candidate Generation
Based on the job posting, generate a fictional "ideal candidate" resume. This represents what the hiring manager imagines as the perfect fit.
Include in the ideal candidate:
- Relevant competitor/industry experience
- All required skills at proficiency level
- Impressive quantified achievements
- Perfect career trajectory for the role
Present to User:
## Ideal Candidate Profile
Based on this job posting, here's what the hiring manager's
"dream candidate" might look like:
[Generated ideal candidate resume summary]
This helps us understand what we're competing against.
Phase 3: Gap Analysis
Compare the user's resume against the ideal candidate. Create a structured comparison that reveals strengths and areas for improvement.
Comparison Format:
| Area | Ideal Candidate | Your Resume | Analysis |
|------|-----------------|-------------|----------|
| Industry Experience | Competitor A, B | Similar Industry C | Transferable |
| Core Skills | X, Y, Z | X, Y | Highlight Z experience |
| Achievements | 50% revenue growth | Project completion | Quantify impact |
Output:
- Strengths: Areas where user matches or exceeds ideal
- Opportunities: Gaps that can be addressed with existing experience
- Limitations: Genuine gaps to acknowledge
Phase 4: High-Spec Version Generation
This phase uses the high-spec-generator subagent.
Why a Subagent? The subagent operates without seeing the original resume details, only:
- Company names
- Job titles
- Employment periods
- The job posting
This isolation prevents anchoring to the original resume's framing, producing fresh perspectives on how achievements could be presented.
Invoke the Subagent:
Use your environment's agent invocation method. In Claude Code, use Task tool with subagent_type="resume-coach:high-spec-generator".
IMPORTANT: Do NOT include original resume content in the prompt. Only provide:
- User's company names, titles, dates
- Target job posting
- Request: Generate a competitive-level resume
This isolation prevents anchoring to the original resume's framing.
Phase 5: Expression Discovery
Compare the high-spec version against the original resume. Help the user identify expressions and framings they can legitimately adopt.
Present Discoveries:
## Expressions from High-Spec Version
The high-spec version uses these compelling framings:
1. "Drove 30% increase in team efficiency"
→ Do you have any similar efficiency improvements?
2. "Led cross-functional initiative spanning 3 departments"
→ Did you work across teams? How many stakeholders?
3. "Implemented data-driven decision framework"
→ Any analytical approaches you introduced?
Use AskUserQuestion: For each compelling expression, ask if the user has similar experiences they haven't highlighted.
Phase 6: Final Resume Assembly
Synthesize all collected information into the final resume:
- From Phase 1: Specific details from recruiter-perspective questions
- From Phase 3: Identified strengths to emphasize
- From Phase 5: Validated expressions and framings
- Original tone: Maintain the user's authentic voice
Quality Criteria:
- All numbers are factual (from user's answers)
- No fabricated experiences
- Expressions are adapted, not copied verbatim
- Maintains user's professional voice
- Directly addresses job posting requirements
Output Options:
- Display in chat as formatted markdown
- Save to file (offer path options)
AskUserQuestion Integration
This skill heavily uses AskUserQuestion for interactive coaching. Follow these patterns:
Fallback: If AskUserQuestion is unavailable in your environment, present one question at a time as a numbered list and wait for user response before proceeding to the next question.
For Recruiter Questions (Phase 1):
- header: Role/scope clarification
- options: Mutually exclusive choices about responsibilities, scope, or metrics
For Expression Validation (Phase 5):
- header: Achievement confirmation
- options: Whether user has similar experiences to adopt the framing
Progress Tracking
Use your environment's task tracking tool to track progress through phases. In Claude Code, prefer TaskCreate/TaskUpdate.
- Phase 1: Recruiter Perspective Questions
- Phase 2: Ideal Candidate Generation
- Phase 3: Gap Analysis
- Phase 4: High-Spec Version (via subagent)
- Phase 5: Expression Discovery
- Phase 6: Final Resume Assembly
Update task status as each phase completes.
Language Handling
- Support both Korean and English
- Match the user's input language
- For bilingual job postings, ask user's preference
- Technical terms may remain in English regardless of language choice
Output Format
Final Resume Structure
# [Name]
## Contact
[Email] | [Phone] | [LinkedIn] | [Location]
## Summary
[2-3 sentences highlighting key value proposition aligned with job posting]
## Experience
### [Job Title] | [Company]
[Start Date] - [End Date]
- [Achievement with metric from user's answers]
- [Responsibility aligned with job requirements]
- [Project highlighting relevant skills]
### [Previous Position]
...
## Skills
[Skills matching job posting requirements]
## Education
[Relevant education]
Subagent Reference
This skill uses one subagent:
high-spec-generator: Generates a competitive-level resume version using only company names, titles, and dates. Located in agents/high-spec-generator.md.
Tips for Effective Coaching
- Don't rush: Each phase builds on the previous one
- Validate everything: Never assume or fabricate details
- Respect user's voice: Adapt framings, don't impose them
- Focus on alignment: Always tie back to job requirements
- Be honest about gaps: Some gaps are real and that's okay
Error Handling
Missing resume: "이력서를 먼저 공유해주세요. 파일 경로나 텍스트를 직접 붙여넣어도 됩니다."
Missing job posting: "어떤 포지션에 지원하시나요? 채용공고 URL이나 내용을 공유해주세요."
Incomplete answers: If user skips questions, note the gap and proceed. Missing context will be reflected in final output quality.