Files

T

Antigravity Agent c24165b8bc refactor(topics): 멀티 에이전트용 지식 재편 — _Common(공통 기본기) + Domain_* 구조

에이전트 8종(대화형/프로그래머 C·S/디자이너/설계자/기획자/QA/PD/PM)에게
[공통 기본 능력 + 롤별 Specialty] 2층으로 지식을 주입하기 위한 재분류.
문서 내용·포맷은 무수정, 폴더 이동만 (6,372개 문서 수 보존 확인).

- Topic_Programming → Domain_Programming (내부 구조 보존)
- Topic_Graphic → Domain_Design
- Topic_Business → Domain_Product
- Topic_General → Domain_General
- _Common 신설: Math(구 Topic_Math_Specialty), Reasoning(구 General/From_Thinking & Reasoning),
  Reasoning_Creativity(구 General/From_창의성), Communication(Poetic_Blog_Writing + From_writing)
- 타 도메인의 From_* 폴더는 유지 (출처 표기일 뿐, 이미 도메인에 맞게 분류된 문서)
- 빈 폴더 정리 (memory/procedures)
- 에이전트→폴더 매핑은 workspace의 .astra/agent-knowledge-map.json (9개 에이전트)

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

2026-07-11 11:05:56 +09:00

5.7 KiB

Raw Blame History

id, title, category, status, canonical_id, aliases, duplicate_of, source_trust_level, confidence_score, verification_status, tags, raw_sources, last_reinforced, github_commit, tech_stack

title

Problem Solving Test (PST)

매 한 줄

"매 McKinsey legacy paper test — 26 case-style multiple-choice questions in 60 minutes — replaced by Solve / Imbellus game in 2018-2019, but the underlying skills (structured thinking, MECE, hypothesis-tree, data interp) 매 still core to consulting interviews". The test is gone; the discipline 매 not.

매 핵심

매 history

1990s-2017: McKinsey PST = pen-and-paper screening before 1st round interview.
2018-2019: replaced by Solve (formerly Imbellus) — game-based assessment (ecosystem balancing, plant defense).
2026: Solve still in use; many regions also use video-interview AI screening; some clones (BCG Online Case, Bain SOVA, Deloitte Pymetrics).
The PST format 매 still copied by other firms / business schools.

매 question types (PST format)

Word problems: "Client X has revenue $Y, costs $Z..."
Data interpretation: read chart/table → infer.
Logic / reading: assumption identification, what-if.
Math: %, ratios, breakeven, growth rates.

매 core consulting frameworks (the actual skill)

MECE: Mutually Exclusive, Collectively Exhaustive — no overlap, no gap.
Issue tree: hierarchical decomposition of the problem.
Hypothesis tree: tree where each node is a falsifiable hypothesis.
80/20 / Pareto: focus on biggest drivers first.
Profit tree: Profit = (P - VC) * Q - FC; decompose each.
Market sizing: top-down (population × penetration × ARPU) vs bottom-up.

매 modern Solve game

6 mini-games: ecosystem, redrock study, plant defense, etc.
Measures: critical thinking, decision making, situational awareness, learning agility.
AI-graded; ~70 min total.

💻 패턴

MECE issue tree (text representation)

Problem: "Why is profit declining?"
├── Revenue down
│   ├── Volume down
│   │   ├── Fewer customers
│   │   └── Less per customer
│   └── Price down
│       ├── Discounts up
│       └── Mix shift to cheaper SKUs
└── Costs up
    ├── COGS up (input price, supplier, waste)
    └── SG&A up (headcount, marketing, IT)

Hypothesis tree

Root hypothesis: "Profit decline is driven by margin compression in Region A."
├── H1: input costs in A rose >5% YoY  [test: supplier invoices]
├── H2: A introduced discounting in Q3   [test: pricing data]
└── H3: A's mix shifted to low-margin SKUs [test: SKU-level P&L]

Profit decomposition

Profit = (Price - VariableCost) × Quantity - FixedCost
        ↓ each is a lever; trace YoY delta to isolate cause

Market sizing (top-down)

US coffee shop market size:
≈ 330M people
× 60% coffee drinkers
× 250 cups/year average
× $4 average cup
≈ $198B  (sanity-check vs $100B reported → adjust assumptions)

Data-interp checklist (PST-style)

Read title + axes + units first.
What is the question actually asking?
Eliminate clearly wrong options (often 2 quickly).
Compute only what's needed (not all entries).
Beware base-rate fallacy: % vs absolute.

Hypothesis-driven case interview opening

"Before diving in, I'd like to structure my thinking.
Given the goal is [restate], I see three areas to explore:
  1. ...  2. ...  3. ...   [MECE check]
I hypothesize the answer lies in [#2] because [reason].
I'd like to start by asking about [data needed to test #2].
Does that approach work?"

Quick math drills (PST training)

- Mental %: 17% of 240 → 10% (24) + 5% (12) + 2% (4.8) ≈ 40.8
- Growth: $100 → $150 over 5 yrs → 1.5^(1/5) ≈ 1.085 ≈ 8.5% CAGR
- Breakeven: FC / (P - VC)
- Return on investment: ΔProfit / Investment

매 결정 기준

상황	Approach
Preparing for McKinsey 2026	Practice Solve (game), not legacy PST
BCG / Bain interview	Online case + structured case method
Improve general problem-solving	MECE + hypothesis tree drills
Ambiguous business problem	Issue tree → 80/20 → hypothesis test
Quantitative case (sizing)	Top-down + bottom-up cross-check

기본값: MECE issue tree → hypothesize → data → synthesize. Format-agnostic.

🔗 Graph

부모: Problem_Solving
변형: McKinsey-Solve · Case-Interview
응용: Strategy-Consulting
Adjacent: MECE · Hypothesis Tree · Pareto-Principle

🤖 LLM 활용

언제: structured business analysis, case interview prep, strategic decomposition of ambiguous problems. 언제 X: technical engineering decisions (use systems thinking instead), pure math optimization.

❌ 안티패턴

Boiling the ocean: not 80/20 — analyze every branch equally.
Non-MECE buckets: overlapping or missing categories.
Hypothesis without test: "I think X" w/ no falsification plan.
Data dump: charts without "so what".
Memorizing PST questions: format is gone since 2019.

🧪 검증 / 중복

Verified (McKinsey careers website, Minto Pyramid Principle, Case in Point textbook).
신뢰도 A.

🕓 Changelog

날짜	변경
2026-05-08	Phase 1
2026-05-10	Manual cleanup — PST history + transferable structured thinking skills

5.7 KiB Raw Blame History Unescape Escape