跳至正文
OpenAgentSkill
浏览已开放 · 目录记录日期 2026-11-21 · 自动更新和账户服务暂时关闭。
搜索与研究

mckinsey-market-research-deck

End-to-end playbook for producing a top-tier, McKinsey-style market-research deck (HTML page-turning presentation + print-ready PDF) for a brand or product category. Covers the research methodology (TAM/SAM/SOM bottom-up, Good/Better/Best framework, competitor profiling, customer pain points, unit economics, business case), the locked McKinsey visual design system, a reusable Python deck engine, an adversarial verify workflow for decision-grade numbers, an image-generation handoff, and a full QC checklist. Use when the user asks to "do market research", build a "market research deck / report", a "McKinsey-style deck / presentation", a "GBB / Good-Better-Best analysis", "market sizing", "competitive landscape deck", "投资/商业案例 deck", or to turn research into a polished slide deck or PDF.

norahe0304-art · 已有静态检查记录

安装与使用

运行命令前先阅读来源与依赖说明。复制命令不会自动安装或执行。

npx skills add norahe0304-art/30x-mckinsey-research-deck --skill mckinsey-market-research-deck

技能说明


name: mckinsey-market-research-deck description: > End-to-end playbook for producing a top-tier, McKinsey-style market-research deck (HTML page-turning presentation + print-ready PDF) for a brand or product category. Covers the research methodology (TAM/SAM/SOM bottom-up, Good/Better/Best framework, competitor profiling, customer pain points, unit economics, business case), the locked McKinsey visual design system, a reusable Python deck engine, an adversarial verify workflow for decision-grade numbers, an image-generation handoff, and a full QC checklist. Use when the user asks to "do market research", build a "market research deck / report", a "McKinsey-style deck / presentation", a "GBB / Good-Better-Best analysis", "market sizing", "competitive landscape deck", "投资/商业案例 deck", or to turn research into a polished slide deck or PDF. allowed-tools:

  • Read
  • Write
  • Edit
  • Bash
  • Glob
  • Grep
  • WebSearch
  • WebFetch
  • Agent
  • Workflow

McKinsey-Style Market-Research Deck

Build a research-backed, visually elite, page-turning deck (HTML reviewed on screen → PDF for sharing). This skill is the distilled, reusable playbook. Read the four reference files as you reach each phase — do not try to hold all of it in head at once.

  • references/methodology.md — how to do the research and what each section must contain
  • references/design-system.md — the locked visual contract (tokens, page types, layout laws)
  • ../mckinsey-deck/assets/deck_engine.py — the canonical engine (owned by the mckinsey-deck style skill; this skill consumes it — never fork a local copy, that's how drift starts)
  • references/qc-checklist.md — the self-verify pass before delivery
  • references/image-handoff.md — the template that hands product/cover images to an image generator

The 7-page spine (always)

  1. The Answer — one answer_slide() right after the cover: the governing thought (the full recommendation in one sentence) + 3–4 pillar conclusions with key numbers. Pyramid Principle: the answer comes first; the rest of the deck is its proof. Drafted in Phase 1.5, finalized last.
  2. Market Overview — size, growth, channel, the structural shift
  3. Brand Landscape — Good/Better/Best ladder + brand-by-brand profiles
  4. Product Categories — per-subcategory competitor price ladder + pain points + the brand's lineup
  5. Customer Pain Points — sourced failure modes, each one a selling-point opening
  6. Opportunities — pain points → product direction
  7. The Solution — positioning, pricing/packaging, the line plan, and the decision pages (bottom-up market sizing, economics, business case) + the thesis

End with a full source register (every URL, numbered).

Workflow (run in order)

Phase 0 — Scope + Day-1 hypothesis

Get: the brand, the parent retailer/company, the category, the geography, the SKU-count target, and the strategic question (usually "what line should we build and why"). Confirm the deck is the deliverable (pure market research), not a precursor needing first-party data. Then write the Day-1 hypothesis — a one-paragraph draft of the answer ("we believe X because A/B/C") before researching. It steers the research (80/20: go deep only on the branches that confirm or kill it) and it is there to be falsified, not defended — revise it whenever the evidence disagrees, and say so in the deck.

Phase 1 — Research → one data file

Do the research per references/methodology.md. Land everything in a single <brand>-data.json (the deck is data-driven from it). Every number must carry a sourceUrl. Schema in methodology.md. Use WebSearch/WebFetch; capture competitor prices/plan tiers live with the capture date (shelf price for goods, plan/ACV for software, cost-to-adopt for OSS/service). Source bar (full rules in methodology.md § The source bar): prices from the vendor's own page only; market sizes from named research, never an SEO aggregator alone; pains quoted verbatim from a named venue; load-bearing inputs need 2 sources or an explicit "judgment call" label; floor of ≥1.5 unique URLs per content page with ≥50% primary/named-research — and zero padding URLs.

Phase 1.5 — Ghost deck (dot-dash storyline)

Before rendering a single page, write the headline-only outline: every page as one action-title sentence, in order, plus a one-line sketch of its exhibit. Then run the horizontal-logic test: read the headlines top to bottom — they must read as one persuasive essay (SCQA arc: situation → complication → question → answer). If a headline doesn't advance the argument, the page gets cut or merged now, before any layout work is spent. Draft the §0 governing thought + pillars here too.

Phase 2 — Generate the deck

Copy the canonical engine ~/.claude/skills/mckinsey-deck/assets/deck_engine.py into the project (single source of truth — engine fixes go back to that file, never to a project-local fork). Point it at <brand>-data.json, set BRAND, compose the 6-section build() (the engine ships the renderers + an example build). Render:

python3 deck_engine.py                       # writes <Brand>-Deck.html
"/Applications/Google Chrome.app/Contents/MacOS/Google Chrome" --headless --disable-gpu \
  --no-pdf-header-footer --print-to-pdf="<Brand>-Deck.pdf" "<Brand>-Deck.html"

After EVERY build, assert structure: div diff must be 0 (an unclosed div breaks pagination).

python3 -c "h=open('<Brand>-Deck.html').read();print('div diff:',h.count('<div')-h.count('</div>'))"

Phase 3 — Decision pages (adversarial verify)

The three pages that turn "opportunity scan" into "decision deck": bottom-up market sizing (TAM/SAM/SOM), economics (validate the value/margin claim with the buildup that fits the category — landed COGS for goods, CAC/payback for SaaS, adoption→conversion for OSS), business case (investment, 3 scenarios, payback). The questions are

阅读完整来源 ↗