Claude Code and Codex skills published by @cyenxchen.
Classifies an idea as a feasibility spike, bounded change, or architectural design and scales the ceremony accordingly. Spike and bounded paths stop after their result. Architectural work can hand off to implementation only after the document is approved and the user separately confirms starting.
Install for Claude Code:
npx skills add cyenxchen/brainstorm-skills --skill brainstorm --agent claude-code --globalInstall for Codex:
npx skills add cyenxchen/brainstorm-skills --skill brainstorm --agent codex --globalTo install the same skill for both hosts:
npx skills add cyenxchen/brainstorm-skills --skill brainstorm --agent claude-code codex --globalThese commands install at user scope. Omit --global to install only in the current project.
What it does:
- Classifies each request as spike, bounded, or architectural and lets you override the classification
- Asks one question at a time to clarify your idea
- Routes text choices through permitted Claude Code or Codex question tools, including async; follows host policy for plain-text approvals and unanswered questions
- Uses optional native feedback throughout design and document review; preserves user-required approvals and actual permission boundaries
- Returns a recommendation for spikes, an in-chat design for bounded changes, or a local
docs/brainstorm/YYYY-MM-DD-<topic>-design.mddraft for architectural work; commits only when authorized - Architectural only: an explicit keep changes
Status: DrafttoStatus: Approved; after verifying that write, the skill asks separately whether to implement now - Starts implementation in the same active task only after an explicit yes; No stops, cancellation/empty/timeout never starts work, and an async delivery acknowledgement remains pending
Tool availability does not guarantee a popup. See question routing and version notes for the observed Codex Default-mode restrictions.
When to use:
When you want to explore feasibility or design a change before implementation, while keeping document approval and implementation authorization distinct.
The brainstorm skill is extracted from obra/superpowers (MIT licensed) and modified to:
- Stop after spike and bounded results, and after Draft or declined architectural outcomes
- Add an explicit architectural Draft-to-Approved state transition and separate same-task implementation handoff
- Remove cross-skill dependencies that don't exist outside the full Superpowers plugin
- Rename to
brainstormto distinguish it from the originalbrainstormingskill
Original author: Jesse Vincent (@obra). The Superpowers methodology this skill comes from is documented in his original release announcement.
MIT — see LICENSE.