lateralus-brainstorm
Skill3 starsChanged 3 months agoMITView on GitHub
Structured brainstorming skill for features, architecture, and ideas. Activates automatically during plan mode. Generates ideas in three tiers: Grounded (practical, shippable), Balanced (assumption-questioning, loosely feasible), Wild (speculative, constraint-free). Use when user invokes /lateralus-brainstorm, says "let's plan", "brainstorm", "what are our options", "how should we approach this", or agent enters plan mode. Use for new work, not debugging.
The file
We split this skill into the rules it gives an agent
12 rules. About 0 are common, 11 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈11
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 27
Check for CLAUDE.md / AGENTS.md / GEMINI.md at repo root — read fully if present.
- Very specific: line 28
If absent: detect stack from go.mod, package.json, requirements.txt, Cargo.toml, etc.
- Distinctive: line 31
No codebase accessible → skip to Phase 2 immediately.
- Distinctive: line 39
Ask before generating anything:
- Distinctive: line 51
If the user's message already makes the horizon obvious, infer it and confirm in one line before generating.
- Distinctive: line 53
Follow-up questions to ask (adapt to the chosen mode):
- Distinctive: line 66
If resources and time were unlimited, what would you build?
- Distinctive: line 76
Always label the tier. Always output the context block before ideas.
- Distinctive: line 142
Ask which idea or direction to develop further.
- Distinctive: line 144
If the user picks something → hand off to planning / implementation, do not generate another batch unprompted.
- Distinctive: line 152
Don't blend modes in one output.
- Distinctive: line 154
If the user says "just give me ideas" without context → ask for goal and horizon first, one question.
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Generate ideas, not fixes. Widen the solution space before narrowing. Always establish goal and horizon first.
Activation
This skill activates in two ways:
- Explicitly: user invokes
/lateralus-brainstorm - Automatically: agent enters plan mode (user says "let's plan", "plan this out", "thinking through", "brainstorm", "what are our options", "how should we approach this")
In plan mode: run Phase 1 silently, then proceed to Phase 2 before producing any plan.
Phase 1 — Read context silently (always first)
Do this before presenting any options:
- Check for
CLAUDE.md/AGENTS.md/GEMINI.mdat repo root — read fully if present. - If absent: detect stack from
go.mod,package.json,requirements.txt,Cargo.toml, etc. SkimREADME.md(first 60 lines). Rungit log -10 --oneline. - Identify: existing architecture patterns, tech constraints, recent directions, open issues or TODOs.
No codebase accessible → skip to Phase 2 immediately.
Surface relevant context briefly before the mode menu (e.g. "I can see this is a React app with a REST backend — that shapes the ideas.").
Phase 2 — Establish goal and horizon
Ask before generating anything:
What are we brainstorming, and what's the horizon?
[1] Grounded — practical ideas you could ship soon, within current constraints
[2] Balanced — ideas that question some assumptions but stay loosely feasible
[3] Wild — constraint-free, blue-sky, no wrong answers
Pick a number, or describe what you're exploring and I'll route you.If the user's message already makes the horizon obvious, infer it and confirm in one line before generating.
Follow-up questions to ask (adapt to the chosen mode):
Grounded:
- What exists already that this builds on?
- What are the hard constraints? (timeline, stack, team size, backwards compat)
- What does success look like — how will you know an idea is good?
Balanced:
- What assumptions are you treating as fixed that might not be?
- Is there a version of this that solves the same goal differently?
- How long does this need to last — prototype, product, or platform?
Wild:
- If resources and time were unlimited, what would you build?
- What would you do if the current stack didn't exist?
- What problem is underneath the problem?
Cold user → one question at a time. Engaged user → batch the mode's questions.
Phase 3 — Generate ideas by mode
Always label the tier. Always output the context block before ideas.
Context block (always first)
Mode: Grounded | Balanced | Wild
Goal: <one line>
Horizon: shipping soon | exploring | blue-sky | unknown
Constraints: <list or "none stated">
Existing context: <key facts from Phase 1, or "none">
Success signal: <what a good idea looks like>Grounded ideas
5-8 concrete, actionable ideas that fit within stated constraints.
Format:
Idea: <name>
What: <one line description>
Why it fits: <constraint alignment>
First step: <smallest thing to try or validate>Optimise for: shippability, familiarity, low risk. No moonshots.
Balanced ideas
5-7 ideas that challenge one assumption each while remaining loosely buildable.
Format:
Idea: <name>
What: <one line description>
Assumption challenged: <what this questions>
Loose feasibility: <why it's not crazy>
Open question: <what needs answering before committing>Optimise for: assumption surfacing, creative-but-credible. Not pure speculation.
Wild ideas
5-6 speculative, constraint-free ideas. State clearly these are not literal plans.
Format:
Idea: <name>
Reframe: <one or two sentences — make it vivid>
Seed: <what grain of this could be extracted into something real>Optimise for: jolting new associations, not implementation readiness. Short. Punchy. No over-justification.
After generating
Always end with:
- Ask which idea or direction to develop further.
- Offer to switch modes: "Want Grounded takes on any of the Wild ideas? Or go deeper on one?"
- If the user picks something → hand off to planning / implementation, do not generate another batch unprompted.
Rules
- Never generate ideas before establishing goal and horizon.
- Wild ideas are labeled "non-literal" — never present them as plans.
- Don't blend modes in one output. Each generation pass is one mode only.
- In plan mode: run this skill automatically, but keep the menu step — don't skip straight to ideas.
- If the user says "just give me ideas" without context → ask for goal and horizon first, one question.
- Repeat passes only when the user asks for more or switches mode.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

