Lateralus brainstorm caveman

Skill SunnyKTuladhar/lateralus/.github/skills/lateralus-brainstorm-caveman

Caveman-compressed brainstorming skill. Same as lateralus-brainstorm — goal and horizon first, three modes (Grounded, Balanced, Wild) — but ~60% fewer tokens. Use when context window is nearly full or you want faster responses.From its SKILL.md

Install

npx -y skills add SunnyKTuladhar/lateralus --skill lateralus-brainstorm-caveman

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

26 days oldThe repository was created 26 days ago. New is not bad, but a brand new repository carrying a familiar-sounding name is the shape a typosquat arrives in, and there has been no time for anyone else to find a problem with it.
1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its file declares

Copied from the file, not written here

The file declares its own license as MIT. That is the author’s claim about this one file, and it is not the same thing as the license GitHub reports for the repository, which is listed with the other numbers below.

SKILL.md

2.5 KB, 569 tokens by cl100k_base, as published. Nobody here has run it

Ideas not fixes. Widen before narrowing. Goal + horizon first, always.

Activation

Explicit: /lateralus-brainstorm-caveman Auto: plan mode ("let's plan", "brainstorm", "what are our options", "thinking through")

In plan mode: read context silently, then present menu.

P1 — Read context silently

Check AGENTS.md / CLAUDE.md / GEMINI.md at repo root — read fully if present.
If absent: detect stack from go.mod / package.json / etc., skim README.md (60 lines), run git log -10 --oneline.
Note: architecture patterns, tech constraints, recent directions.

No codebase → skip to P2.

P2 — Goal + horizon

Surface context briefly if found. Then:

Brainstorming — pick a mode:

  [1] Grounded  — practical, shippable soon, within constraints
  [2] Balanced  — question one assumption, stay loosely feasible
  [3] Wild      — constraint-free, blue-sky

Pick a number or describe what you're exploring.

Horizon obvious from message → infer + confirm in one line, proceed. Distressed / deadline → suggest [1] first.

Follow-up (batch if user is engaged, one at a time if cold):

[1] Grounded: What exists already? Hard constraints? What does a good idea look like? [2] Balanced: What assumption is treated as fixed but maybe isn't? Prototype or platform? [3] Wild: Unlimited resources — what would you build? Problem beneath the problem?

P3 — Generate

Context block first, always:

Mode: Grounded | Balanced | Wild
Goal / Horizon / Constraints / Existing context / Success signal

Grounded — 5-8 ideas: Idea: <name> | What: <one line> | Why it fits: <constraint> | First step: <smallest test>

Balanced — 5-7 ideas: Idea: <name> | What: <one line> | Assumption challenged: <what> | Open question: <what to answer>

Wild — 5-6 ideas (non-literal — state this upfront): Idea: <name> | Reframe: <1-2 sentences> | Seed: <extractable real grain>

Rules

Never generate before goal + horizon established. Wild = non-literal. Never present as plan. One mode per pass. No blending. End: ask which direction to develop, offer to switch modes.