Auto-triggered skill state machine
Work is decomposed into reusable skills and triggered by conversation state so the agent clarifies constraints, produces a readable spec, then moves into an executable plan and implementation.
Superpowers is a workflow-first skills library for coding agents. Instead of relying on ad-hoc prompting, it hardens the fragile parts of software work—spec clarification, task breakdown, execution cadence, TDD discipline, and review loops—into skills that trigger automatically. You describe the goal, it aligns on a readable spec, produces an executable plan, and then advances in checkpoints so agent output stays engineering-grade rather than improvisational code.
| ✕Traditional Pain Points | ✓Innovative Solutions |
|---|---|
| Agents often skip clarification and acceptance criteria, jump into code, and create rework and context drift. | Superpowers encodes the design-to-execution loop as auto-triggered skills, replacing prompt improvisation with process guarantees. |
| Parallel agent work tends to lose cadence and review gates, so tasks sprawl and become hard to finish cleanly. | Batch execution plus two-stage review (spec compliance → code quality) keeps parallel work aligned and shippable. |
1/plugin marketplace add obra/superpowers-marketplace1/plugin install superpowers@superpowers-marketplace1/help1/plugin update superpowers| Core Scene | Target Audience | Solution | Outcome |
|---|---|---|---|
| Team delivery from spec to ship | engineering teams and tech leads | standardize clarification, spec sign-off, planning, and execution cadence | less rework and more reliable delivery |
| Controlled parallel execution | projects running parallel work | push tasks in batches with explicit review gates | parallelism without sprawl, with auditable progress |
| TDD discipline for risky changes | engineers maintaining critical systems | enforce test-first and verification steps through skills | shift risk left and reduce regressions |
