🚨 BREAKING: Someone just gave Claude Code a complete software development brain.
It's called Superpowers and it makes Claude plan, test, and ship code without going off the rails.
No junior dev babysitting. No context loss. No hallucinated plans.
100% Opensource.
Superpowers is an agentic skills framework that gives Claude Code a full engineering workflow out of the box:
- Brainstorming (spec before code)
- Implementation planning (bite-sized tasks)
- Subagent-driven development (parallel execution)
- TDD enforcement (RED-GREEN-REFACTOR, no shortcuts)
All triggered automatically. You don't configure anything.
Here's what makes it different:
Most AI coding setups let Claude jump straight to writing code.
That's where everything goes wrong.
Superpowers forces Claude to:
→ Ask what you're actually building
→ Write a spec you can read in chunks
→ Build a plan clear enough for a junior dev to follow
→ Test before claiming it works
Structure first. Code second.
Technical setup:
- Works as a Claude Code plugin (one-line install)
- Also supports Codex and OpenCode
- Skills trigger automatically based on context
- Includes 15+ composable skills (TDD, debugging, git worktrees, parallel agents)
- MIT license
/plugin marketplace add obra/superpowers-marketplace
/plugin install superpowers@superpowers-marketplace
That's it.
Real-world workflow:
You say: "Build me a Stripe payment integration"
Superpowers:
1. Asks clarifying questions → locks in spec
2. Breaks work into 2-5 min tasks with exact file paths
3. Launches subagents per task with two-stage code review
4. Enforces failing test → passing test → commit
Claude works autonomously for hours. Without derailing.
Why this matters:
AI coding agents fail because they skip the thinking.
They write code before understanding the problem.
They declare success before running tests.
They hallucinate plans and barrel through anyway.
Superpowers solves this with mandatory workflows, not suggestions.
The agent checks for relevant skills before any task. Every time.
The skills library alone is worth the install:
Testing → RED-GREEN-REFACTOR enforcement
Debugging → 4-phase root cause process (no guessing)
Brainstorming → Socratic design refinement
Parallel agents → Concurrent subagent workflows
Git worktrees → Parallel dev branches
Code review → Pre and post review checklists
Each skill is a composable module. Add what you need.
Being honest about limitations:
- Best results with Claude Code (Codex/OpenCode support is newer)
- Subagent-driven dev can get expensive on complex tasks
- Early plugin marketplace ecosystem (still growing)
- Works best when you actually follow the brainstorm → plan → execute order
But it's open source, so the community is actively fixing all of this.
If you use Claude Code and you're not using Superpowers, you're doing vibe coding with extra steps.
This is what structured AI development actually looks like.
100% Opensource:
What's killing your current Claude Code workflow — the planning phase or the testing phase?github.com/obra/superpowe…
I hope you've found this thread helpful.
Don't forget to bookmark for later.
Follow me @heyrimsha for more.
If you enjoyed reading this post, please support it with like/repost of the post below 👇
Share this Scrolly Tale with your friends.
A Scrolly Tale is a new way to read Twitter threads with a more visually immersive experience.
Discover more beautiful Scrolly Tales like this.