Now liveThe Skillselion MCP - thousands of ranked skills, loaded into your agent mid-task. No install.Get it →

obra/superpowers

15 skills2M installs4M starsGitHub

Install

npx skills add https://github.com/obra/superpowers

Skills in this repo

1BrainstormingExplore and refine ideas into fully formed designs through collaborative dialogue. Ask clarifying questions, propose alternatives, present design sections for approval, and write spec documents before beginning implementation.299kinstalls2Systematic DebuggingSystematic debugging methodology for finding root cause before attempting any fix. Four-phase process: root cause investigation (reproduce, gather evidence, trace data flow), pattern analysis (find working examples, identify differences), hypothesis testing (form theory, test minimally), and implementation (create test, fix, verify). Emphasizes architectural questioning after multiple failed fixes.203kinstalls3Writing Planswriting-plans is a Superpowers meta-skill for generating implementation plans before any code is written. The skill assumes zero context about the codebase and produces bite-sized tasks with exact file paths, test examples, and commit boundaries. Developers use writing-plans when they have a spec or requirements for multi-step work and need to decompose it into a reviewable, checkpointed plan.200kinstalls4Using Superpowersusing-superpowers is the meta-skill at the heart of obra's Superpowers library: it forces the agent to discover and follow relevant skills before improvising. The workflow is strict — if a skill might apply, invoke it immediately with the Skill tool, even when the task feels simple or faster to answer from memory. It maps Claude Code tool names to platform equivalents (Task to spawn_agent on Codex, etc.) without rewriting skill content. The skill covers brainstorming, TDD, debugging, plans, and code review as a coordinated process library rather than one-off prompts. Developers reach for it when they want repeatable engineering discipline instead of ad-hoc chat answers. It complements individual Superpowers skills by governing when and how they load, making the whole plugin behave like a senior engineer who always checks the playbook first.198kinstalls5Requesting Code ReviewA pre-commit code review checklist that identifies issues by severity and blocks on critical problems. Integrates into Superpowers workflow between implementation tasks to catch bugs and quality issues before merging.182kinstalls6Test Driven DevelopmentRED-GREEN-REFACTOR test-driven development methodology enforcing failing tests before implementation. No production code without a failing test first - mandatory for features, bugs, and refactoring with zero exceptions.180kinstalls7Executing PlansExecuting-plans is a superpowers skill that systematically executes written implementation plans with critical review. It loads a plan, reviews it for concerns, executes all tasks step-by-step, and reports completion status. Use when you have a detailed plan and need structured task execution with verification checkpoints.168kinstalls8Subagent Driven DevelopmentSubagent-Driven Development is a Superpowers workflow skill for executing implementation plans inside the current session by dispatching a fresh implementer subagent per task, a spec-and-quality reviewer after each task, and a broad whole-branch review at the end. Controllers construct isolated context per subagent using task-brief and review-package scripts so diffs and reports stay in files instead of polluting session history. The process mandates continuous execution without pausing between tasks unless blocked, ambiguous, or complete, and requires explicit model selection per role to balance cost and judgment. Implementer statuses DONE, DONE_WITH_CONCERNS, NEEDS_CONTEXT, and BLOCKED each have defined handling paths, while reviewers must receive brief, report, and diff file paths without pre-judged severity instructions. A durable progress ledger at .superpowers/sdd/progress.md prevents re-dispatching completed tasks after context compaction. Required companion skills include using-git-worktrees, writing-plans, requesting-code-review, and finishing-a-development-branch, with test-driven-development for implementers.159kinstalls9Verification Before CompletionEnforces evidence-based completion claims before commits or PRs. Requires running verification commands and confirming output before making any success assertions. Eliminates unverified status claims by gating completion on fresh evidence.159kinstalls10Receiving Code ReviewRespond to code review feedback with technical rigor instead of performative agreement. Requires restating requirements, verifying against codebase reality, and implementing with technical evaluation. Eliminates blind implementation of unclear feedback.151kinstalls11Writing SkillsTest-driven approach to creating and editing skills. Treats skill authoring as documentation TDD: write pressure scenarios with agents, verify failures without skill, write the skill, verify compliance. Ensures skills teach techniques agents actually need.150kinstalls12Dispatching Parallel AgentsDelegate independent tasks to specialized parallel agents with isolated context. Each agent handles one problem domain without inheriting session state. Faster than sequential investigation for multiple unrelated issues.148kinstalls13Using Git WorktreesUse native worktree tools or git worktrees for isolated workspaces. Detect existing isolation first, prefer platform-native tools, fall back to manual git worktrees only when needed. Never fight the harness.147kinstalls14Finishing A Development BranchDecides whether to merge, create a PR, keep a branch, or discard after task completion. Verifies test status, runs cleanup, and presents clear options based on worktree state. Integrates with git-based development workflows in Superpowers.145kinstalls15SuperpowersSuperpowers is a core skills library for Claude Code covering test-driven development, debugging, collaboration patterns, and other proven techniques. It packages reusable engineering workflows so the agent applies disciplined methods consistently. Aimed at developers who want a strong baseline of agent capabilities out of the box.0installs

This week in AI coding

Five minutes, every Monday - the tools, releases and tactics for developers.

unsubscribe anytime.