Superpowers is a software development method for coding agents. It is a set of composable skills (brainstorming, planning, TDD, debugging, subagent execution, code review, branch finishing) plus a bootstrap skill that makes the agent check for and use them before acting. It is for developers who run agents in Claude Code, Codex, Cursor, Gemini and other harnesses.
Why read it
Nearly every skill ends with an Excuse/Reality table built from things agents were seen telling themselves. Each tempting shortcut gets a short rebuttal. Because the excuse is named ahead of time, the agent can spot the thought in itself as a warning sign. The tables don't need to restate the rule.
writing-skills sorts failures into types and gives each a matching form. Rule violations get prohibitions plus red flags. Output with the wrong shape gets a positive recipe. A missing element gets a required slot in the template. It reports that prohibitions made shaping problems worse in wording tests, and it warns against nuance clauses, which reopen the rule for negotiation.
Skill descriptions start with "Use when..." and list only the conditions that should trigger the skill. The stated reason is that an agent that reads a summary of the workflow will follow the summary and skip the skill body.
The execution skills keep progress in a git-ignored ledger file. Its first line names the plan it belongs to, and each task gets one completion line. After compaction the agent is told to trust the ledger and git log over its own memory. This prevents a costly failure the skill says it saw in real sessions: re-running work that was already finished.
When the agent hits a conflict or ambiguity, it decides and keeps going. It records each decision in a fixed format that includes the reason and the cost if wrong, and lists every ruling in its final message. The human keeps visibility without being interrupted.
Subagents get a task brief extracted to a file and hand their full reports back as files. Pasting session history into dispatches is forbidden, backed by a measured example. The skill also says to set the model explicitly for each role, because an omitted model quietly inherits the most expensive one.
The core non-negotiable rule is set apart in a fenced block or HARD-GATE tag, along with a line saying that breaking the letter is breaking the spirit. The gate spells out what counts as "building" and names its only exceptions, which leaves little room for reinterpretation.
verification-before-completion pairs each kind of claim with the evidence it needs and the evidence that isn't enough. One example: an agent reporting "success" is not proof; the VCS diff is. diagnosing-superpowers applies the same idea by requiring a path:line citation for every finding.
A model read each skill to find these. Top to bottom is the usual order. Select a box to read the skill, or a line to see the text behind it.
diagnosing-superpowers, dispatching-parallel-agents, receiving-code-review
Every one is a skill.
| SKILL | STARTED BY | DESCRIPTION | LINKS |
|---|---|---|---|
| brainstorming | You or the agent | You MUST use this before any creative work - creating features, building components, adding functionality, modifying behavior, or planning anything new, in software or out of it (a talk, a business, a renovation). | 3 |
| diagnosing-superpowers | You or the agent | Use when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | 0 |
| dispatching-parallel-agents | You or the agent | Use when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | 0 |
| executing-plans | You or the agent | Use when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | 9 |
| finishing-a-development-branch | You or the agent | Use when implementation is complete, all tests pass, and you need to decide how to integrate the work | 1 |
| receiving-code-review | You or the agent | Use when receiving code review feedback, before implementing suggestions, especially if feedback seems unclear or technically questionable - requires technical rigor and verification, not performative agreement or blind implementation | 0 |
| requesting-code-review | You or the agent | Use when completing tasks, implementing major features, or before merging to verify work meets requirements | 1 |
| subagent-driven-development | You or the agent | Use when executing implementation plans with independent tasks in the current session | 6 |
| systematic-debugging | You or the agent | Use when encountering any bug, test failure, or unexpected behavior, before proposing fixes | 2 |
| test-driven-development | You or the agent | Use when implementing any feature or bugfix, before writing implementation code | 1 |
| using-git-worktrees | You or the agent | Use when starting feature work that needs isolation from current workspace or before executing implementation plans - ensures an isolated workspace exists via native tools or git worktree fallback | 4 |
| using-superpowers | Automatic | Use when starting any conversation - establishes how to find and use skills, requiring skill invocation before ANY response including clarifying questions | 8 |
| verification-before-completion | You or the agent | Use when about to claim work is complete, fixed, or passing, before committing or creating PRs - requires running verification commands and confirming output before making any success claims; evidence before assertions always | 1 |
| writing-plans | You or the agent | Use when you have a spec or requirements for a multi-step task, before touching code | 5 |
| writing-skills | You or the agent | Use when creating new skills, editing existing skills, or verifying skills work before deployment | 5 |
From obra/superpowers at 2ce8aa1, 2026-10-21. Skills read by DeepSeek V4.1 Flash on 2026-11-04. Summary and techniques by Claude Opus 5.5 on 2026-11-06.