Use before any creative architecture, feature design, or stack selection - explores requirements, generates 2-3 approaches, and uses TypeSafe Jev (jev-axi pick) to converge on the optimal technical decision
Scanned 9/22/2026
Install to Claude Code
npx -y skills add AkashPriyadarshii/jev-superpowers --skill jev-brainstorming --agent claude-codeInstalls into .claude/skills of the current project.
Are you the author of Jev Brainstorming?
Add the live security badge to your README — it updates automatically with every re-scan.
[](https://www.skillsdirectory.com/skills/akashpriyadarshii-jev-brainstorming)More formats (shields.io, HTML) on the badges page.
---
name: jev-brainstorming
description: Use before any creative architecture, feature design, or stack selection - explores requirements, generates 2-3 approaches, and uses TypeSafe Jev (jev-axi pick) to converge on the optimal technical decision
---
# Jev Brainstorming: Systematic Design with Typed Convergence
Inherits all requirements exploration and collaborative dialogue rules from `superpowers:brainstorming`, but eliminates ungrounded LLM guessing on architectural tradeoffs using TypeSafe Jev System One.
## The Workflow
### 1. Classify Scope
- **Spike**: Quick feasibility question. 2-3 sentences in chat.
- **Bounded**: Modifying an existing, readable flow. Short design in chat.
- **Architectural**: New subsystem, stack choice, or major refactor. Full process with written spec.
### 2. Divergent Exploration
- Ask the 1-2 clarifying questions that matter most.
- Identify the core tension (latency vs simplicity, SQLite vs Postgres, sync vs async).
- Present 2-3 viable technical approaches with honest pros and cons.
### 3. Jev Convergence Gate (MANDATORY)
Before recommending an approach or asking the user to decide, run `jev-axi pick` to evaluate the architectural decision deterministically:
```bash
jev-axi pick "<Tradeoff Question>" \
--options "<opt1>,<opt2>,<opt3>" \
--text "<Context, constraints, user persona, latency & memory targets>"
```
**Evaluation Rules:**
1. **Report Live Jev Output**: State the picked option, its probability distribution, and calibrated confidence score.
2. **Confidence Threshold**:
- `Confidence >= 0.80` (Band: `act`): Present as the strongly recommended choice.
- `Confidence < 0.80` (Band: `confirm`/`escalate`): Explain why the tradeoff is close and explicitly ask the human partner to break the tie.
3. **Ponytail Check**: If one option is standard library or already-installed dependency, bias the context toward Rung 3/5.
### 4. Human Approval Gate
Present the Jev-scored recommendation in chat. Do NOT touch code or write implementation plans until the human partner says "yes" or selects an alternative.
## Failure Modes
See docs/CONFIDENCE.md for thresholds. When the gate tool is missing, the key is invalid, the registry is offline, or confidence falls below the Stop band: STOP, state which input failed, and never degrade to unverified guessing silently.
Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.
No comments yet. Be the first to comment!