Activate a 4-stage coding discipline framework that forces Claude to plan before coding, isolate changes on a branch, write tests first, and self-review outp...
Coding
Right-Size Superpowers
Try itScale Superpowers workflow depth to task risk
What it does
Use when the user explicitly wants Superpowers workflow depth calibrated to task scope and risk, especially when routine or bounded work is receiving heavyweight process.
The skill document
Right-Size Superpowers
Preserve correctness while scaling Superpowers process to what can materially change the result. The user's requested outcome and constraints remain authoritative.
Choose the Smallest Sufficient Route
| Route | Observable condition | Response |
|---|---|---|
| Direct | Outcome is clear, scope is bounded, action is reversible or read-only, and no material design choice is unresolved | Act immediately |
| Focused | Root cause is unknown, several related files are involved, or regression risk is meaningful | Use only the one discipline needed to resolve that risk |
| Full | Architecture, public interfaces, persistent data, security, destructive actions, or materially ambiguous product behavior are involved | Explain the trigger and use the appropriate deeper workflow |
For Direct and Focused work, do not automatically chain brainstorming, writing-plans, worktrees, TDD, subagents, or review workflows merely because code is involved. Domain-specific skills required for the artifact or tool still apply.
Execution Contract
- Infer routine details from the request and existing project conventions.
- Inspect only the context needed to identify the target and avoid collateral changes.
- Make the smallest coherent change that completes the request.
- Verify with the closest existing check or targeted test once. Broaden verification only after a failure, a cross-cutting change, or evidence of wider risk.
- Add a regression test when it can catch a meaningful behavior failure. Do not add tests that merely mirror a reversible configuration, copy, wording, or formatting change.
- Review the resulting diff for scope, accidental edits, and unmet requirements.
- Report the outcome, changed files, verification performed, and any unresolved risk concisely.
Ask one focused question only when its answer would materially change the result. Do not request another approval for reversible work already authorized by the user.
Example
Request: “Rename this setting and update its two references.”
Use Direct: locate the definition and references, make the rename, run the nearest relevant check, inspect the diff, and report completion. Do not create a design document, implementation plan, new worktree, subagent review, or unrelated test suite.
Red Flags
- Treating every code change as architectural work
- Producing a plan artifact for a mechanical edit
- Re-running unchanged checks without new evidence
- Loading unrelated process skills “just in case”
- Asking for approval after the user already authorized a reversible action
If one appears, return to the smallest sufficient route.
Related skills
用AI写代码一时爽，项目越写越乱火葬场？Superpowers方法论5道Gate卡住质量：设计没想清楚不准动手、没写测试不准提交。把「能跑就行」升级成工程级交付，让AI帮你写出可维护的代码。支持全栈项目（Web/移动端/API/数据）、团队协作规范、CI/CD集成指南。从需求到上线的完整工程化流水线，杜绝"AI生...
3 installs
用AI写代码一时爽，项目越写越乱火葬场？Superpowers方法论5道Gate卡住质量：设计没想清楚不准动手、没写测试不准提交。把「能跑就行」升级成工程级交付，让AI帮你写出可维护的代码。支持全栈项目（Web/移动端/API/数据）、团队协作规范、CI/CD集成指南。从需求到上线的完整工程化流水线，杜绝"AI生...
4 installs
Automates repetitive tasks using scripts, scheduled jobs, and workflows. Streamlines build/deploy processes, cron/Task Scheduler tasks, and multi-step manual operations. Reduces human error and saves time by handling routine actions reliably. Ideal for automating any scriptable procedure or schedule
3 installs
Deep Coding is a multi-agent development system designed for complex software projects. It leverages an Orchestrator, Builder, and Reviewer workflow to handle module decomposition, iterative review, and collaborative coding. Activate when users request deep coding, multi-agent collaboration, or comp
1 stars
Gunakan saat user menghadapi task yang butuh reasoning dalam, analisis kompleks, problem-solving kreatif, sintesis long-context, atau planning multi-step. Menyediakan framework kognitif + self-correction loop untuk meningkatkan kualitas output pada task tersebut. Aktif saat user minta 'pikir lebih dalam', 'analisis mendalam', atau 'rencana multi-langkah'.