dx-brainstorm
Skill8 starsChanged 4 months agoMITView on GitHub
Think through a raw idea before committing to it — weigh alternatives, price doing nothing, and route to a change, an effort, or nothing at all.
In this repository21 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- dx-archiveSkill
- dx-diagnoseSkill
- dx-distillSkill
- dx-domain-discoverSkill
- dx-domainSkill
- dx-frameSkill
- dx-implementSkill
- dx-initSkill
- dx-lessonSkill
- dx-newSkill
- dx-planSkill
- dx-refactor-discoverSkill
- dx-referencesSkill
- dx-researchSkill
- dx-reviewSkill
- dx-review-triageSkill
- dx-roadmapSkill
- dx-standards-discoverSkill
- dx-standards-updateSkill
- dx-tddSkill
The file
We split this skill into the rules it gives an agent
4 rules. About 0 are common, 2 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 20
Before converging — unless §1 found this already shipped, which settles the ramp:
- Distinctive: line 24
The riskiest assumption and the cheapest test of it: If a one-day spike settles it, that spike may be the whole change.
- Very specific: line 33
Stop when one ramp fits, every unknown that would *change the routing* is resolved, and the user signals enough.
- Very specific: line 42
Ramps 1–2: write nothing and never load that reference.
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 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
dx-brainstorm
Runs before dx-new on an idea nobody has yet decided is worth building. Concluding that nothing should be built is a success, not a run that failed to produce work.
Guards. context/ unscaffolded (changes/ and efforts/ absent — empty is normal) → stop, say to run /dx-init. No user available to answer → stop; don't invent the appetite and constraints you came here to ask about. And "this is simple enough to just do it now" is the red flag, not the shortcut — whether it's simple enough is the conclusion, not the bypass.
1 — Sources before questions
A question whose answer already sits in the repo is homework, not conversation. Climb in order, stopping as soon as something answers: foundation/glossary.md + lessons.md → context/standards/ → context/changes/** and context/archive/** (already decided, shipped, or rejected?) → the codebase → the user.
2 — Diverge
Before converging — unless §1 found this already shipped, which settles the ramp:
- Restate the problem without the user's proposed solution. It arrives pre-shaped; the problem underneath is usually wider or narrower.
- ≥2 genuinely different alternatives, plus a priced do-nothing. Different mechanism, not the same idea at two sizes. Price what the pain costs per week and who absorbs it. Strawmanning it is the failure mode.
- The riskiest assumption and the cheapest test of it. If a one-day spike settles it, that spike may be the whole change.
- Refuse vague framing. "Improve X" — push to an observable difference, or there's nothing to build.
- The bundling test — would each capability function without the other? Yes → effort. No → change. It only decides ramps 3 vs 4.
- "What breaks if this succeeds?"
3 — Converge
Invoke dx-references with topic interview — one question at a time, and run step 5's adversarial pass before presenting the conclusion, since routing is hard to reverse once a container exists. A critical objection goes back to the user as a question; don't self-resolve it.
Stop when one ramp fits, every unknown that would change the routing is resolved, and the user signals enough. Routing-sufficiency, not exhaustiveness — the rest belongs to dx-frame and dx-plan.
4 — Take one of four exit ramps
- Nothing worth building — no container, no artifact, no
Next:line. State the conclusion and why the do-nothing won. - Already covered by an open or archived container — cite its path and stop.
- One shippable unit →
context/changes/<id>/change.md(invokedx-referenceswith topicchange-md) +brainstorm.md, whose shape is${CLAUDE_SKILL_DIR}/references/brainstorm-md.md— read it before writing. Stamptype— normallyfeature; it gates the plan's characteristics, so decide it rather than leaving it to be guessed. - ≥2 independently shippable capabilities →
context/efforts/<id>/effort.md(invokedx-referenceswith topiceffort-md) +brainstorm.md, same shape. Identity file only — noroadmap.md, though the reference describes one, because the slices aren't yours to cut. Efforts carry notype; don't invent one.
Ramps 1–2 write nothing and never load that reference; when the rejection is load-bearing, offer /dx-lesson — the one durable trace either leaves. Ramps 3–4: propose the <id> so the user can redirect it (kebab-case, space-free).
Tie-breakers: don't manufacture work to have a handoff. Already shipped → ramp 2, never ramp 1. Appetite is the user's call. If naming the affected areas took real work, that's ramp 4 over 3.
Not a chain
Creating the container is this skill's deliverable; the no-auto-chain rule binds the step after it. Never run /dx-plan, /dx-frame, or /dx-roadmap. Slices stay dx-roadmap's, child changes dx-new's.
Done when
One ramp taken, its artifacts (if any) written, the outcome printed. The left-hand label selects the case — print only what follows it, plus any caveat the adversarial pass left standing. Then stop:
No user: <what couldn't be asked — and that no appetite or constraint was invented>
Nothing to build: <one-line conclusion — why the do-nothing won>
Already covered: Covered by <path> (status: <status>)
One change: Brainstorm written: context/changes/<id>/brainstorm.md
Next: /dx-frame <id> → /dx-plan <id> (frame optional)
An effort: Brainstorm written: context/efforts/<id>/brainstorm.md
Next: /dx-frame <id> → /dx-roadmap <id> (frame optional)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.

