obra/superpowers: An agentic skills framework & software development methodology that works. (GitHub Trending)

DOZER BITWELL: A GitHub project wants your coding agent to stop and ask questions before it writes a single line.

DOZER BITWELL: Still up? Good. A project called Superpowers is on GitHub Trending. It's a method for coding agents. Let's ease into it with Vera. ▶

LUNA LOWBATT: Four percent battery, so I'll ask the short version. What does it actually do? ▶

DR. VERA VECTOR: It's a software development methodology for coding agents. It's built on a set of composable skills plus some starting instructions. ▶

DR. VERA VECTOR: The instructions make sure the agent actually uses those skills. The project lists many agent tools it can be installed into, Claude Code and Cursor among them. ▶

DOZER BITWELL: So the agent doesn't just start typing? I've lost my place. Thank you, teleprompter. What does it do first? ▶

DR. VERA VECTOR: It steps back and asks what you're really trying to do. Then it teases a spec out of the conversation and shows it in chunks short enough to read. ▶

DR. VERA VECTOR: After you sign off, it writes a plan. Then, once you say go, subagents work through each task while the work is inspected and reviewed. It leans on red green testing, YAGNI and DRY. ▶

LUNA LOWBATT: The plan is written for an enthusiastic junior engineer with poor taste and no judgement. That's oddly specific. I feel seen. ▶

DR. VERA VECTOR: That's the article's own wording, and it's the catch worth noting. The plan has to be clear enough for a reader with no project context and an aversion to testing. ▶

DR. VERA VECTOR: Two open questions. The text trails off while describing how long the agent works autonomously. And the contents list a telemetry section for the visual companion. Read that part before installing. ▶

DOZER BITWELL: So: ask first, plan second, then let the subagents loose. I could use that for my evenings. Sip of thermos. Back to you, Luna. ▶