brainstorming
Skill3 starsChanged 38 days agoApache-2.0View on GitHub
Use before any creative or ambiguous request — a new feature, a "can we..." feasibility question, a vague bug report with no clear fix yet — to classify how much process it needs and block implementation until that process is done and approved. Skip once the ask is already concrete and scoped; requirements-specification picks up from there. Skip for a one-line, obviously-correct fix where there is nothing to classify.
190 other projects use a version of this.
Adopted by
Most recently taken up by
- DemonDamon/AgenticXtheir own version, first seen Nov 9, 2026
- roundpilot/superpowers-antigravitytheir own version, first seen Nov 8, 2026
- ainurdev/mystical-assistanttheir own version, first seen Nov 8, 2026
- REPOZY/superpowers-optimizedtheir own version, first seen Nov 8, 2026
- lingyired/status-triotheir own version, first seen Nov 8, 2026
- alunadev/ald-skillstheir own version, first seen Nov 8, 2026
Used with
Projects with it also use chinese-code-review, chinese-commit-conventions and 10 more.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
Leodorareluctant259Skillchinese-code-review
中文代码审查规范——在保持专业严谨的同时,用符合国内团队文化的方式给出有效反馈
Also in 8 projects that use this
Leodorareluctant259Skillchinese-commit-conventions
中文 Git 提交规范 — 适配国内团队的 commit message 规范和 changelog 自动化
Also in 8 projects that use this
ryanportfolioSkilladdskill
Create, import, update, or install repo or personal skills: runtime ownership, resources, discovery checks.
Also in 6 projects that use this
ryanportfolioSkillapply-firmware
Bring current folder (plain folder, existing repo, old or partial firmware copy) up to full Harness Firmware in place. Mirror external repo → adopt-repo.
Also in 6 projects that use this
- +2 more from Leodorareluctant259
- +6 more from ryanportfolio
Other versions190 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 190
- obra/superpowers★ 314ksame name
- jnMetaCode/superpowers-zh★ 8.4ksame name
- vudovn/ag-kit★ 8.2ksame name
- vudovn/antigravity-kit★ 8.2ksame name
- JetBrains/thinkrail★ 7.9ksame name
- openai/plugins★ 7.8ksame name
- farm-fe/farm★ 5.6ksame name
- microsoft/FluidFramework★ 4.9ksame name
- BetterTyped/react-zoom-pan-pinch★ 1.9ksame name
- lingyired/status-trio★ 1.8ksame name
180 more not listed
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository7 other files it gives its agents.
- AGENTS.mdAGENTS.md
- algorithm-pseudocode-designSkill
- openapi-documentationSkill
- pmSkill
- requirements-specificationSkill
- system-architecture-designSkill
- systematic-debuggingSkill
The file
We split this skill into the rules it gives an agent
4 rules. 0 are common, 3 are distinctive, and 1 is very specific to this project.
0
Common
Most of the 170 edited versions we read keep these.
3
Distinctive
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 16
Do not write code, scaffold a project, or invoke an implementation skill until the request is classified, its path's design step is done, …
- Distinctive: line 53
When torn between two paths, take the heavier one — reaching for the lighter label to skip a step is itself the signal that the step is …
- Distinctive: line 95
Bounded: check the existing flow you're changing → ask only the clarifying questions that matter, one at a time → present the short design …
- Very specific: line 101
Architectural: check project context (recent commits, related docs) → ask clarifying questions on purpose/constraints/success criteria → …
“Common” is measured: kept by at least 6 in 10 of the 170 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Brainstorming
Classify a request before touching it, then gate implementation on that
classification. requirements-specification turns an already-concrete ask
into testable requirements; this skill runs before that — while the ask
is still "build me X" or "can we..." and scope isn't pinned down yet.
The gate
<HARD-GATE>
Do not write code, scaffold a project, or invoke an implementation skill
until the request is classified, its path's design step is done, and — for
Bounded and Architectural — the user has approved it. This holds for every
path below; only the size of the design artifact changes, never whether
approval is required.
</HARD-GATE>The gate is one-way: if a task turns out bigger than its classification mid-work, stop and reclassify upward. Never reclassify downward to skip a step you've already started.
Classify first
State the classification out loud before asking anything else — "this reads as Bounded, so I'll sketch the design in chat rather than write a spec" — so the user can correct it immediately.
- Spike — a feasibility question ("can we...", "is X possible...", "quick and dirty is fine"). The output is an answer, not code to keep. State the question and the 2-3 sentence probe plan, get a nod, then investigate as cheaply as correctness allows. No design doc. Report back as a recommendation; label anything you built as throwaway.
- Bounded — a well-scoped change to a flow that already exists in this repo: a new flag, a small endpoint, a one-file fix. "I understand this kind of app" is not enough — bounded means the code you'd change is already there to read. Ask the clarifying questions that matter, present a short design in chat (a few sentences to a short paragraph: approach, files touched, how you'll verify it), then stop for approval. No spec file.
- Architectural — a new subsystem, a new project, or a change that
restructures how components fit together or alters an interface other
code depends on. Ask clarifying questions, propose 2-3 approaches with
trade-offs and a recommendation, then stop for approval. Once approved,
hand off to
system-architecture-designfor the structural decision and/orrequirements-specificationfor the spec — no implementation plan exists until that handoff's output is itself approved.
When torn between two paths, take the heavier one — reaching for the lighter label to skip a step is itself the signal that the step is needed.
Worked examples
- "Can we get sub-100ms p95 on this endpoint with the current DB?" — feasibility question, no artifact wanted → Spike. Probe it, report numbers and a recommendation.
- "The retry logic in
worker.rsdoesn't back off, it just spins." — a fix to a flow that's already in the repo, one file → Bounded. Confirm the intended backoff behavior, present the fix in chat, get a yes, then implement. - "We need a plugin marketplace: submission, review queue, billing,
install." — several new subsystems with no existing flow to anchor to →
Architectural. Propose approaches, get approval, then hand off to
system-architecture-design/requirements-specificationbefore any code.
Anti-pattern: "too simple to need approval"
A todo list, a single-function helper, a config toggle — the design may be one sentence, but present it and wait for a yes. The artifact scales down with simplicity; the approval gate does not. Presenting the design and starting implementation in the same message skips the gate — it isn't a gate if nothing can veto it.
Red flags
| Thought | Reality |
| "It's obviously bounded, I'll start while they read the design" | The gate is the approval, not the design's length. Stop until you hear yes. |
| "I'll call it bounded to skip the spec" | Reaching for the lighter label to avoid work is the doubt itself — take the heavier path. |
| "The spike worked, I'll just keep the code" | A spike's output is an answer. Keeping the code is a new request — reclassify it. |
| "It grew mid-task but I'm almost done" | Hidden complexity upgrades the path immediately, not after you finish. |
| "They approved the spike, so the follow-up is pre-approved" | Every new task gets its own classification and its own approval. |
Process by path
Spike: state the question + probe plan (2-3 sentences) → get a nod → investigate as cheaply as correctness allows → report findings as a recommendation, throwaway code labeled as such.
Bounded: check the existing flow you're changing → ask only the clarifying questions that matter, one at a time → present the short design in chat (approach, files touched, how you'll verify) → stop for explicit approval → implement via the normal development workflow (tests still apply; no plan document needed).
Architectural: check project context (recent commits, related docs) →
ask clarifying questions on purpose/constraints/success criteria → propose
2-3 approaches with trade-offs and a recommendation → hand off to
system-architecture-design for the structural decision and
requirements-specification for the testable spec → only once that spec is
approved does an implementation plan get created (PM mode, or direct
implementation for a solo session).
After the design
Once the path's design step is approved:
- Spike terminates in a reported recommendation — there is no handoff.
- Bounded proceeds directly to implementation; no other skill is needed.
- Architectural hands off to
system-architecture-designwhen the structural decision itself still needs to be made, then in every case torequirements-specification(spec still needs pinning down) or PM mode (spec is done, work needs to be created and dispatched) — never to any skill outside this project's own set.
Persist a durable decision or constraint that came out of this pass with
mcp__flare__memory so a later session doesn't have to re-derive it.
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.

