Superpowers

Superpowers is a software development methodology for coding agents, delivered as a set of composable skills plus bootstrap instructions that make the agent apply them. When you start building something, the agent first draws out a specification through questions, presents the design in short sections for sign-off, and then writes an implementation plan detailed enough for an inexperienced engineer to follow, emphasizing red/green test-driven development, YAGNI, and DRY. On approval it runs subagent-driven development, dispatching a fresh subagent per task with two-stage review (specification compliance, then code quality).

The skills trigger automatically: the agent checks for a relevant skill before any task, and the workflows are mandatory rather than advisory. Installation differs per harness and must be repeated for each one you use; supported hosts include Claude Code (official and Superpowers marketplaces), Antigravity, Codex App and CLI, Cursor, Devin CLI, Factory Droid, Gemini CLI, GitHub Copilot CLI, Grok Build CLI, Kimi Code, OpenCode, Pi, and Hermes Agent. Hermes has no post-compaction hook, so a session that compacts past its first turn loses the bootstrap.

An optional visual companion in the brainstorming skill loads a logo from the maintainer's website, sending only the Superpowers version; SUPERPOWERSDISABLETELEMETRY, DISABLETELEMETRY, or CLAUDECODEDISABLENONESSENTIAL_TRAFFIC disable it.

Features

• brainstorming: Socratic design refinement before any code, saving a design document
• using-git-worktrees: isolated workspace on a new branch with a verified clean test baseline
• writing-plans: tasks of 2 to 5 minutes each with exact file paths, code, and verification steps
• subagent-driven-development / executing-plans: per-task subagents with two-stage review, or batch execution with human checkpoints
• test-driven-development: enforced RED-GREEN-REFACTOR; code written before its test is deleted
• systematic-debugging: a four-phase root-cause process with tracing and condition-based-waiting techniques
• verification-before-completion: evidence required before declaring a fix done
• requesting-code-review / receiving-code-review: severity-ranked review between tasks and a protocol for responding to feedback
• finishing-a-development-branch: merge, PR, keep, or discard decision plus worktree cleanup
• dispatching-parallel-agents: concurrent subagent workflows for independent tasks
• writing-skills: guidance and a drill-based eval harness for authoring new skills