Brainstorm
Install this before creative or structural decisions so the agent runs a Double Diamond brainstorm grounded in memory instead of jumping straight to implementation.
Overview
Brainstorm is a journey-wide agent skill that facilitiates Double Diamond ideation with memory-backed grounding—usable whenever a solo builder needs to explore options before committing.
Install
npx skills add https://github.com/hyperb1iss/hyperskills --skill brainstormWhat is this skill?
- Double Diamond rhythm: Ground → diverge on problem → converge on definition → diverge on solutions → converge on plan
- Explicit split: AI favors divergent volume; humans own convergent judgment—phases are guidance, not a rigid script
- Persistent memory via Sibyl to avoid re-litigating solved problems across sessions
- Mined from 100+ real production brainstorming sessions
- Activates on brainstorm, ideate, architecture decisions, new features, and “how should we approach” language
- 100+ real brainstorming sessions
- Double Diamond multi-phase shape (Ground through Converge on plan)
Adoption & trust: 565 installs on skills.sh; 13 GitHub stars; 2/3 security scanners passed (skills.sh audits).
What problem does it solve?
You face an ambiguous feature, architecture, or product direction and risk shipping the first idea the model suggests without structured exploration.
Who is it for?
Solo builders at any journey phase who want facilitated ideation before specs, code, or launch commitments.
Skip if: Situations where requirements are already approved and you only need execution or mechanical refactors—skip brainstorming and invoke implementation-focused skills instead.
When should I use this skill?
Before creative work—brainstorm, ideate, design session, explore options, new feature, new project, architecture decision, or design exploration.
What do I get? / Deliverables
You leave with a grounded problem definition, evaluated solution options, and a converged plan ready for specs, tickets, or a downstream planning skill—without redoing old debates thanks to persistent memory.
- Grounded problem definition
- Shortlisted solution options with rationale
- Converged plan or next-step brief for spec/planning skills
Recommended Skills
Journey fit
Useful at every journey phase - explore requirements and options before committing to a direction.
Where it fits
Explore audience problems and solution directions before picking a niche to validate.
Diverge on MVP cuts and converge on a prototype boundary before writing code.
Facilitate an architecture decision session before the agent drafts an implementation plan.
Brainstorm positioning and channel experiments before committing launch copy.
Re-ground on incident or UX pain, diverge fixes, then converge on the next iteration theme.
How it compares
Use instead of unstructured chat brainstorming when you want Double Diamond phases, explicit human/AI roles, and memory that remembers prior decisions.
Common Questions / FAQ
Who is brainstorm for?
Indie and solo builders who want a repeatable ideation facilitator for features, architecture, and project inception across the whole product journey.
When should I use brainstorm?
At Idea when researching what to build; at Validate when scoping or prototyping forks; at Build before architecture or major features; at Ship/Launch when exploring rollout or positioning options—whenever commitment is still reversible.
Is brainstorm safe to install?
It is process guidance that may use persistent memory (Sibyl); review the Security Audits panel on this page and your memory store policies before enabling in production repos.
SKILL.md
READMESKILL.md - Brainstorm
# Collaborative Brainstorming Structured ideation using the Double Diamond model, grounded in persistent memory. Mined from 100+ real brainstorming sessions across production projects. **Core insight:** AI excels at divergent phases (volume, cross-domain connections). Humans excel at convergent phases (judgment, selection). Separating the two, and using Sibyl to avoid re-exploring solved problems, is the shape that consistently produces useful brainstorms. **How to read this skill:** the phases below describe the natural rhythm of a good brainstorm, not a procedure to march through. Skip phases that don't apply. Revisit earlier phases when new info changes the frame. Use judgment about when to compress, when to skip to action, and when divergent exploration is actually warranted. ## The Shape ```dot digraph brainstorm { rankdir=TB; node [shape=box]; "1. GROUND" [style=filled, fillcolor="#e8e8ff"]; "2. DIVERGE: Problem" [style=filled, fillcolor="#ffe8e8"]; "3. CONVERGE: Define" [style=filled, fillcolor="#e8ffe8"]; "4. DIVERGE: Solutions" [style=filled, fillcolor="#ffe8e8"]; "5. CONVERGE: Decide" [style=filled, fillcolor="#e8ffe8"]; "EXIT → Any skill" [style=filled, fillcolor="#fff8e0"]; "1. GROUND" -> "2. DIVERGE: Problem"; "2. DIVERGE: Problem" -> "3. CONVERGE: Define"; "3. CONVERGE: Define" -> "4. DIVERGE: Solutions"; "4. DIVERGE: Solutions" -> "5. CONVERGE: Decide"; "5. CONVERGE: Decide" -> "EXIT → Any skill"; } ``` --- ## Phase 1: GROUND (Memory-First) Lean on existing knowledge before generating new ideas. The cost of a Sibyl search is low; the cost of re-discovering a pattern we already learned is high. ### Common moves - **Search Sibyl** for related patterns, decisions, and known constraints. Useful queries: `sibyl search "<topic keywords>"`, `sibyl search "<related architecture>"`, plus a quick scan of existing tasks/epics on the topic. - **Surface constraints** that aren't up for debate: tech stack locks, budget, timeline, conventions you'd be foolish to violate. - **Present prior art** before ideating: "Sibyl has 3 relevant entries: [pattern X], [decision Z], [gotcha W]. Want to factor these in?" If Sibyl already has a directly applicable answer, surface it first. The brainstorm is then about whether to apply it as-is, adapt it, or genuinely diverge, not re-deriving it from scratch. --- ## Phase 2: DIVERGE: Explore the Problem Space **Goal:** Generate breadth. Understand what we're actually solving before reaching for solutions. ### Common moves - **Lean toward one load-bearing question at a time.** Stacking five questions buries the signal; building understanding incrementally surfaces what actually matters. The exception is when the user fires multiple parallel asks; then answer them in parallel rather than artificially serializing. - **Reframe the problem** from multiple angles: user view ("as a [user], I need..."), system view ("the system currently..."), constraint view ("we're bounded by..."). - **Spawn parallel Explore agents** when the problem space is genuinely large: research how similar projects solve this, map the existing codebase surface, search for SOTA approaches. The discipline here is staying in problem space when the pull toward solutions is strong. If the problem is already crisp, skip ahead. This phase exists to prevent solving the wrong thing, not to perform exploration. ### Anti-patterns - Jumping to solutions before the problem frame is clear - Stacking questions when the answers depend on each other - Dismissing vague input ("make it faster" is a valid