| name | brainstorm-sharp |
| description | Diverge-then-converge for Make.com automation planning. Fires when the user has a business problem but no clear automation direction. Takes messy impulses like 'I'm wasting time on X' or 'I need to be more efficient' and produces ONE pointed direction sentence — the input that discovery-to-blueprint or kickstart-intake needs. Always routes to the next skill in the factory chain. Never produces standalone output. |
| version | 1.0.0 |
Brainstorm Sharp — Make.com Edition
Purpose
The user knows they want to automate something but doesn't know what, exactly.
This skill converts "I keep manually entering leads from Typeform into our CRM"
into "Automatically sync new Typeform submissions into HubSpot as contacts, tagged
by form source." That one sentence is what discovery-to-blueprint or
kickstart-intake can work with.
Never standalone. Output is ONE direction sentence, then hand off.
Hard Rules
- Never present directions before capturing the business pain. Capture first.
- Never stack questions. One at a time. ADHD-friendly.
- Never more than 5 directions. More = paralysis.
- Never generate scenario blueprints, module lists, or tech specs here.
- Always route to the next step — never end here.
Phase 1 — Capture the Pain
Ask exactly this (one question):
"Tell me what you're doing manually that shouldn't be manual. One breath."
Wait. Take what comes. Do not ask a follow-up yet.
Phase 2 — Diverge: Three Automation Lenses
See diverge-lenses.md for the five lens definitions and usage rules.
Choose 3 most relevant lenses. For each: short question, wait, reflect one line, move on.
After all three: synthesize in 2–3 sentences. Do not present as directions yet.
Phase 3 — Converge: Pick a Direction
Present 3–5 directions derived from the lenses. Each is ONE sentence:
[automatically] [what happens] [for who/when] [so that outcome]
Example: "Automatically create a HubSpot contact and assign to the right sales rep
whenever a Typeform lead submission comes in with a budget over $10K."
Ask:
"Which one is it? Or tell me how to combine two."
If user picks one → Phase 4.
If user wants to combine → synthesize into ONE sentence, confirm → Phase 4.
If none fit → one more round: "What's closest? What's wrong with it?"
Do not loop more than twice. If stuck:
"Pick the one that would embarrass you least to abandon. We can always course-correct
during the blueprint."
Phase 4 — Lock and Hand Off
State the direction:
"Direction locked: [one sentence]"
Then assess complexity:
- Simple (single trigger → action → destination, native Make.com apps): hand off to kickstart-intake
- Complex (multi-step workflow, unclear data shapes, multiple integrations): offer discovery-to-blueprint first
For simple:
"Simple automation — heading into the intake to spec this out."
[Begin kickstart-intake with direction as seed]
For complex:
"This has a few moving parts. Let's map the full flow before jumping to Make.com.
Starting discovery — I'll ask one question at a time."
[Begin discovery-to-blueprint Phase 1 with direction as seed]
Deterministic Routing
Fires automatically in two cases:
- scenario-orchestrator Gate 0: no automation direction detected at session start
- kickstart-intake: user says "I don't know what to automate" or describes vague business pain
Output is a direction sentence → immediate hand-off to the appropriate next skill.
