Superpowers
An agentic skills framework and software-development methodology by Jesse Vincent (
obra) that makes coding agents brainstorm a spec, plan with TDD, and run reviewed implementation loops. Kevin-Wiki retains and locally enforces the useful invariants; the upstream plugin is not currently installed or globally injected by this v3 checkout.
What it is
Superpowers (obra/superpowers) is a complete software-development methodology built from composable skills and bootstrap instructions. In an upstream installation, the skills trigger automatically; that statement describes the plugin's behavior, not Kevin-Wiki's current runtime state. Source: obra/superpowers README, HEAD 44c9b2d6e889982ac18c27d05a19fefe335194e1, reviewed 2026-08-10
Current source snapshot on 2026-08-10: main HEAD 44c9b2d6e889982ac18c27d05a19fefe335194e1, MIT license, 270,249 stars, and 24,158 forks. Upstream now documents eleven harness targets and optional visual-companion telemetry that can be disabled with SUPERPOWERS_DISABLE_TELEMETRY or the supported Claude telemetry flags. Source: captured source + GitHub API, 2026-08-10
The core loop the framework imposes from the moment you start a coding agent: Source: obra/superpowers README, https://github.com/obra/superpowers, 2026-06-12
- Don't jump to code. Step back and ask what you're really trying to do.
- Tease out a spec, then show it back in chunks short enough to actually read.
- After sign-off, write an implementation plan clear enough for "an enthusiastic junior engineer with poor taste, no judgement, no project context, and an aversion to testing" — emphasizing red/green TDD, YAGNI, and DRY.
- On "go," run subagent-driven development — agents work each task, inspecting and reviewing as they go, often running autonomously for a couple of hours without deviating from the plan.
Cross-harness
Superpowers installs per harness (install it separately for each): Claude Code, Antigravity, Codex App, Codex CLI, Cursor, Factory Droid, Gemini CLI, GitHub Copilot CLI, Kimi Code, OpenCode, and Pi. On Claude Code it ships through Anthropic's official plugin marketplace; on Codex App/CLI it ships through the official Codex plugin marketplace. Source: obra/superpowers README, https://github.com/obra/superpowers, 2026-06-30
# Claude Code (official marketplace)
/plugin install superpowers@claude-plugins-official
Where it fits Kevin's stack
The old checkout described the upstream plugin as always-on. The v3 source of truth is narrower: AGENTS.md, the agent operations hub, the review contract, and local iteration/verification skills implement the useful behavior without claiming that upstream hooks are loaded. Current equivalents are:
| Upstream behavior | Current local owner |
|---|---|
| Skills-first routing | AGENTS.md, agent-operations-hub, runtime skill registry |
| Clarify and design before implementation | task-specific workflow plus explicit plan when complexity warrants it |
| Executable implementation loop | agent-iteration-loop, improve, and run-contract proof |
| Systematic diagnosis | specialized debugging/review skills |
| Evidence before completion | doctors, tests, review receipts, and closeout map |
| Parallel/subagent execution | optional only when the user or active environment explicitly authorizes it |
Superpowers remains strong upstream evidence for Kevin's "design before building / prove before completion" discipline and shares the same gate shape as The Eval Loop (Slop Is an Output Problem), Spec-Driven Development, and Harness Engineering. It is not the authority for current agent permissions or subagent policy; live system instructions and the local operating graph are.
The current upstream basic workflow adds a fuller branch lifecycle: brainstorming, using-git-worktrees, writing-plans, subagent-driven-development or executing-plans, test-driven-development, requesting-code-review, and finishing-a-development-branch. Kevin's repo does not need to mirror every upstream skill name; the local invariant is that each serious task has a design/spec, isolated work path when needed, executable plan, review loop, verification proof, and cleanup/merge decision. Source: obra/superpowers README, 2026-06-30
It is also, per the marketing-angle bookmarks, broadly applicable beyond engineering — the brainstorm -> spec -> tasks -> subagents -> review pipeline maps onto any structured knowledge work, not just code. The reviewed image makes this concrete for marketing: brainstorm campaign angles/questions/brief, plan bite-sized deliverables/file paths, execute with researcher/writer/editor/publisher agents, review against brand foundation/rubric, then verify final voice and publish to channels. Source: X/@shannholmberg thread, 2026-03-14; Source: local asset /wiki/assets/x-bookmarks/2032892199751528486/image-01.jpg, 2026-06-30
Timeline
-
2026-08-05 | Preserved the seven-stage operational detail: design approval committed before implementation; consent-gated worktrees; plans without placeholders; per-task spec and quality review; tests written before implementation; severity-ranked review; branch finish; a progress ledger for compaction recovery; and a debugging stop that questions architecture after three failed fixes. Upstream figures remain unverified and no plugin was installed. Source: X/@shannholmberg
2079482908012839397 -
2026-08-10 | Re-reviewed Superpowers through the cross-department skills bookmark and current HEAD
44c9b2d6e889982ac18c27d05a19fefe335194e1. Corrected the v3 runtime claim: the local graph absorbs the method, but this checkout does not install or inject the upstream plugin globally. Added the current optional telemetry boundary and kept subagent execution conditional on active authorization. Source: X/@cyrilXBT, 2026-07-16; Source:obra/superpowers, reviewed 2026-08-10 -
2026-06-30 | Deep-reviewed the Shann marketing thread and local image artifact, then refreshed upstream Superpowers to
v6.0.3/main896224c4b1879920ab573417e68fd51d2ccc9072. Decision: update the page and tool/workflow routing; no new local skill, because Kevin already enforces the method throughAGENTS.md, Agent Operations Skills, Agent Operations Skills, and review/verification skills. Source: X/@shannholmberg, 2026-03-14; Source: GitHubobra/superpowers, 2026-06-30 -
2026-07-04 | Reviewed @iruletheworldmo's March Superpowers bookmark as duplicate adoption evidence. Updated the source snapshot to
v6.1.1; no new executable skill was created because Kevin's local agent rules already enforce the Superpowers-style loop. Source: X/@iruletheworldmo, 2026-03-17; Source: GitHubobra/superpowers, 2026-07-04 -
2026-03-15 | @gavinpurcell: "I'm using the superpowers skill for almost everything I'm working on. it's great." Builder endorsement beyond the dev core. Source: X/@gavinpurcell, 2026-03-15
-
2026-03-14 | @shannholmberg thread reframed Superpowers (83K stars, "trending daily") for non-developers, walking through the brainstorm → spec → tasks → parallel subagents → review pipeline and how to add custom domain skills. 2,855 likes, 7,437 bookmarks. Source: X/@shannholmberg, 2026-03-14
-
2026-06-12 | Page created from X-bookmark absorption (
2032892199751528486,2033046276959346701); methodology and install sourced from the obra/superpowers README and Kevin'sAGENTS.mdenforcement section. Source: x-bookmark-absorb, 2026-06-12