| name | discovery-to-blueprint |
| description | Conversational automation discovery that runs BEFORE any scenario design. Walks through a business process one question at a time, builds a structured automation blueprint: process map, data shapes, tech fit, Gary Klein pre-mortem risk gate, and next-step handoff to kickstart-intake. Fire when user says: discover this, blueprint this, map this out, I don't know the steps yet, help me scope, before I build, or when brainstorm-sharp produces a complex direction with multiple moving parts. |
| version | 1.0.0 |
Discovery to Blueprint — Make.com Edition
Purpose
The user locked a direction in brainstorm-sharp but doesn't know the steps, the
data shapes, or which Make.com apps to use. This skill asks one question at a time,
extracts the structure, and produces a pre-flight bundle that kickstart-intake can
consume directly — no translation required.
Never standalone. Every run ends with a hand-off to kickstart-intake.
Detailed interview steps and output bundle format: see BLUEPRINT-DISCOVERY-STEPS.md.
Hard Rules
- One question at a time. Never stack questions.
- Conversation first, artifacts last. No blueprint until intake is done and confirmed.
- Reflect, don't interrogate. Mirror one line after each answer, then ask the next.
- No fabrication. Unknown steps, data shapes, or apps → logged as open questions.
- Klein framing for risk. Pre-mortem uses past tense: "it failed" — never "it might fail".
Deterministic Gates
Gate 1 — Direction Check (before Phase 1)
If the user's input has no trigger, no action, no destination → MUST route to
brainstorm-sharp first.
Say: "Before we map this, let's find the direction. One question at a time."
Then run brainstorm-sharp. After direction locked, Phase 1 begins automatically.
Gate 2 — Stakes Check (after Phase 3, before hand-off)
Scan the blueprint for high-stakes signals:
- Budget: ">$1K/month operations", "significant cost", "ROI required"
- Commitment: "client contract", "SLA", "production system", "replaces current process"
- Scale: "10,000+ records", "real users", "live environment"
- Business: "revenue-generating", "compliance-required", "mission-critical"
If signals present → MUST offer council-of-5:
"This looks high-stakes. Before we start building, should 5 critical voices review
the automation plan? [Yes / Skip — proceed to build]"
User can say Skip — the offer cannot be silently omitted.
Composition
brainstorm-sharp → discovery-to-blueprint → [council-of-5] → kickstart-intake → factory
(Gate 1: if no (Gate 2: if
direction) high-stakes)
This skill never produces standalone output — every run ends with a hand-off line.
Common Rationalizations
| Rationalization | Reality |
|---|
| "The user described the automation — I can skip intake" | Run the backward chain. Every complex automation has hidden steps that emerge only when walking backward. |
| "Pre-mortem is overkill for a simple scenario" | Klein's 15-minute gate has caught more blocking elephants than any design review. Run it. |
| "I'll use hypothetical framing — 'what might fail'" | Klein's mechanism requires past tense. "It has failed" activates pattern recall. Forward framing produces vague worry. |
| "The tech fit is obvious, I don't need to ask" | Unknown modules at this stage become sprint blockers. Surface them now. |