skip to content
文章 · 2026年6月 Posts · June 2026

superpowers

作者 By Yc Lin·2026年6月27日27 June 2026·4 min read

ai· github

obra/superpowers

Update:2026-06-27 ⭐ Stars：240,000 🍴 Forks：21,300 🔗 GitHub：https://github.com/obra/superpowers

General

In the area of software development, large language models have been widely used to support programming, but traditionAIAssistants tend to generate codes directly, and lack of thinking about the overall architecture, which can lead to test failures and confusion in engineering logic. To address the lack of systematic methodology, blind re-engineering and poor quality of delivery of artificial intelligence when it is self-generated,obra/superpowers♪ As a faceAIThe self-skilled framework for the coded intelligent body and the development of methodologies were developed.

The project builds a pool of skills that can be combined, based on the core idea of “process over speculation” (see para.Skills LibraryIt is bound by the imposition of structured workflows.AIbehavior: Socrates brainstorms to extract demand before code starts, automatically creating isolations after design has been approvedGit, and generate a step-by-step implementation plan with very small particle size.TDD) process, following the “Red-Green-Reconstruct” cycle, even automatically deletes uncertified codes generated before the test is prepared.

Eventually, the project was of various typesAIProgram Client (e.g.,Claude Code、Gemini CLI、GitHub Copilot CLIIt divides complex development tasks into autonomous sub-intelligent-driven development and ensures quality of delivery through multi-stage normative and code quality reviews. The project applies to upgrading.AI”Codification autonomy, guaranteed coverage of large-scale engineering tests and reduced”AIGarbage code (Agentic SlopThe Automated Software Collaboration Site.

Technical Repository

  • Shell Script (51.6%): core life cycle hooks for the preparation of the bottom of the framework, initialization of the automated environment,GitWorking Tree (Worktrees) Management and appropriate matching scripts across the Platform’s tool chain.

  • JavaScript / TypeScript (43.9%): Subject logic, preparation and paragraphs for building pluginsAIProgrammingHarness(e.g.,Claude Code、Codex、Kimi) the extension of the docking and the test operator (s)Runner）。

  • HTML (2.2%): for optional visual partners (in %1Visual Companion) function and local visual telemetry interface.

  • Python (1.9%): Support script preparation, call for partially automated processes and integrated assessment testing.

Scope / Impact

  • Field of application: Applicable to autonomyAgentAutomation of development-driven, agile development teamsTDDProcess landing, multi-intelligent co-programming and enterprise levelAIcoding process quality audit.

  • Reference to organizations/projects: The framework is now in the original support and in depthAnthropicIt’s…Claude Code、GoogleIt’s…Gemini CLI、GitHub Copilot CLIandKimi CodeandOpenCodeWaiting for mainstreamAIEncoding tool. Its development entity isPrime RadiantThe project is currently receiving a high degree of attention in open-source communities (see also the following:240k Stars）。

评论Comments