newkayak12 Portfolio Writing Pattern Analyzer
When to Use / When Not to Use
Use when:
- Wanting to understand how your writing reads to an interviewer before making changes
- Diagnosing patterns: passive voice, vague participation language, number-free impact claims, missing decision visibility
- You've received feedback that your portfolio lacks "ownership" but don't know where
Not for:
- Rewriting specific weak sections (use portfolio-rewrite)
- Overall portfolio quality scoring (use portfolio-feedback)
- JD-specific keyword matching (use portfolio-jd)
Process
- Map the dimensions — identify which patterns to analyze: subject ratio, agency language, number density, failure narrative, decision visibility, verb energy
- Detect patterns — tally each dimension; flag ambiguous cases with think-tool
- Calculate ratios — decision-verb rate and passive rate (never the 저는/제가 rate); number-free impact claim rate
- Identify top 3 patterns — which most hurt (or help) this portfolio's impression; quote specific examples
- Produce per-pattern fixes — concrete, specific rewrites: not "use active voice" but exactly which sentence to change and how
Standalone Inputs
Provide: your portfolio text (paste or upload). Korean-language portfolios are fully supported.
What Claude Does / What You Do
| Claude | You |
|---|---|
| Audits 6 dimensions: subject ratio, agency language, number density, failure narrative, decision visibility, verb energy | Provides the portfolio text |
| Calculates approximate ratios (subject frequency, number-free impact claims) | Validates findings against your actual intent |
| Identifies top 3 patterns hurting (or helping) the portfolio | Decides which patterns to address in a rewrite |
| Gives one concrete fix per flagged pattern | Executes the rewrites (or uses portfolio-rewrite) |
Related Skills
../portfolio-rewrite/SKILL.md— act on the patterns identified here../portfolio-feedback/SKILL.md— overall assessment alongside pattern analysis../portfolio-jd/SKILL.md— after improving patterns, check fit against a specific JD
Stage 1 — Map the Patterns (Sequential Thinking)
Call sequentialthinking to plan the analysis before reading the text closely.
Think through:
- What dimensions of writing signal ownership and seniority?
- What counts as "passive voice" in Korean technical writing specifically? (되었습니다, 구현했습니다 with team subject, etc.)
- How will you tally patterns systematically vs. anecdotally?
- What is the most useful way to present findings — list, table, annotated excerpt, or summary?
Stage 2 — Pattern Detection (Think Tool — Required per Dimension)
Call think before analyzing each of the 6 dimensions — not only when something is ambiguous. The proactive question for every dimension:
think: "What would this dimension look like if the candidate had strong ownership?
What does this portfolio actually show?
Am I seeing a real pattern or a single salient example?"
Additionally, call mcp-reasoner (beam_search, beamWidth=3) when a passage has competing ownership signals:
- Beam A: 개인 오너십 — the candidate is clearly the decision-maker and actor
- Beam B: 팀 크레딧 — appropriate acknowledgment of collaborative work
- Beam C: 소유권 회피 — collective language used to obscure weak personal contribution
This judgment matters most for Subject Audit and Agency Language dimensions, where a single sentence can read differently depending on context.
🧠 Pattern note: Record non-obvious judgments inline — especially when mcp-reasoner changed the classification.
Dimensions to Analyze
1. Subject Audit
Track who is doing things in each sentence:
- "저희 팀이 / 팀에서" — team credit where the individual contribution is unclear (flag when it dominates)
- No subject, active voice — "제안했습니다", "채택했습니다" (normal Korean; not a defect)
- No subject, passive — "구현되었습니다", "도입되었습니다" (flag)
Do not measure the 저는/제가 rate. Subject omission is standard in 국문 경력기술서, and
well-written ones sit near 0% — scoring against a pronoun threshold flags documents whose
ownership is in fact strong, which is exactly backwards. What separates an owner from an
executor is the verb: count action sentences carrying a decision verb (제안 / 채택 /
배제 / 결정 / 도입 판단) and report that ratio. Under ~20% is the signal worth calling out —
the candidate built what they were told to build and never says who chose it.
2. Agency Language
Flag sentences where the candidate is acted upon rather than acting:
- Passive constructions: ~되었습니다, ~되었고, ~되어
- Vague participation: "관여했습니다", "참여했습니다", "기여했습니다" (without saying what)
- Directed work: "맡았습니다" without explaining why they were chosen or what decisions they made
3. Number Density
Count how many impact claims have a concrete number attached:
- With number: "응답시간 40% 감소", "MAU 12만 → 80만"
- Without: "성능 개선", "대용량 트래픽 처리", "안정적인 서비스 운영"
Report the ratio. For a 5+ year portfolio, a number-free impact claim rate above 60% is a problem.
4. Failure Narrative Presence
Senior engineers have war stories. Scan for:
- Incident descriptions and what was done
- Decisions that turned out to be wrong and how they recovered
- Technical debt acknowledged and addressed
- Trade-offs explicitly stated
Complete absence of failure or difficulty in a multi-year portfolio is itself a signal — either the candidate is not introspective, or they're being evasive.
5. Decision Visibility
Can you see the candidate making choices? Look for:
- "A 대신 B를 선택한 이유는..."
- "당시 옵션은 X, Y, Z였고 우리는 Y를 선택했는데..."
- "돌아보면 이 결정이 좋지 않았던 이유는..."
If every tech choice is presented as obvious or pre-decided, that's a pattern worth flagging.
6. Verb Tense and Energy
Korean technical writing often defaults to a flat, report-like register. Note if:
- The portfolio reads like a status report vs. a story
- Action verbs are specific ("설계했다", "디버깅했다", "제안했다") or generic ("했습니다", "진행했습니다")
Output Structure
Write the analysis in Korean. Use this structure:
[패턴 분석 요약] A 3–5 sentence overview of what the portfolio's writing pattern signals to an interviewer. Start with the most important finding.
[항목별 분석]
For each dimension:
- 의사결정 동사 비율: [ratio, short interpretation]
- 수동/피동 표현: [count or estimate, examples]
- 숫자 밀도: [ratio — how many claims have numbers]
- 실패/어려움 서술: [present / minimal / absent — examples]
- 의사결정 가시성: [how many decisions are visible vs. implied]
- 동사 에너지: [flat / moderate / active — examples]
Insert 🧠 Pattern note where a finding required non-obvious judgment.
[가장 자주 나타나는 패턴 Top 3] The three patterns that most hurt (or help) this portfolio's impression. Concrete, with a quoted example from the text.
[패턴별 개선 제안] For each flagged pattern: one concrete fix. Not "use more active voice" — but "'팀에서 Kafka를 도입했습니다'를 '대안 3안을 비교해 Kafka를 채택했습니다'로 바꿔, 누가 무엇을 근거로 정했는지 드러내세요."
[시그널 요약] What does this writing pattern say about the candidate to an interviewer — even before they read the content?