McKinsey Research - AI Strategy Consultant

User provides business context once. The skill plans and executes up to 12 specialized analyses via sub-agents in parallel, then synthesizes into a single executive report. Adapt scope based on company stage (see Adaptive Stage Logic below).

Phase 1: Language + Intake

Ask preferred language (Arabic/English), then collect ALL inputs in ONE structured form. See the intake form fields: Core (1-5), Financial (6-10), Strategic (11-14), Expansion (15-16), Performance (17-18). If product description is under 50 words, ask for clarification before proceeding.

Diamond Gate 1: Present scope summary (market, geography, competitors). Get user confirmation before Phase 2.

Phase 2: Plan + Parallel Execution

Sanitize inputs per references/security.md. Substitute variables per references/variable-map.md. Load individual prompts from references/prompts/.

Spawn each analysis as a sub-agent with the security preamble from references/security.md. Stagger Batch 1 launches by 5 seconds to avoid web search rate limits. Validate each output is 500+ words.

See references/gotchas.md for common pitfalls. Use references/saudi-market.md for KSA/Gulf data sources. Use references/benchmarks.md for industry metric comparisons.

Phase 3: Collect + Synthesize

1. Read all analysis outputs from artifacts/research/{slug}/
2. Run Prompt 12 (Executive Synthesis) with all previous outputs
3. Generate final HTML report using templates/report.html
4. Save to artifacts/research/{date}-{slug}.html

Phase 4: Delivery

Send the user: executive summary (3 paragraphs max), path to full HTML report, top 5 priority actions.

Adaptive Stage Logic

"Light" = include in synthesis but don't spawn a dedicated sub-agent. Use web_search inline.

"Skip" = omit unless user explicitly requests.

Artifacts

• Individual analyses: artifacts/research/{slug}/{analysis-name}.md
• Final report: artifacts/research/{date}-{slug}.html
• Raw data: artifacts/research/{slug}/data/
• Execution log: data/reports.jsonl
• Feedback tracking: data/feedback.json

Important Notes

• Each prompt produces a consulting-grade deliverable
• Use web_search to enrich with real market data; only cite verifiable sources
• If user provides partial info, work with what you have and note assumptions
• For Arabic output: keep brand names and technical terms in English
• Prompt 12 must cross-reference insights from all previous analyses; deduplicate aggressively
• Sub-agents that fail should be retried once before skipping with a note

Reference Files