TIMEWELL
Solutions
Free ConsultationContact Us
TIMEWELL

Unleashing organizational potential with AI

ISO/IEC 27001 (ISMS) certification mark (SGS / ISMS-AC)

ISO/IEC 27001:2022 certified Certificate No. JP26/00000255 Scope: Planning, development and operation of SaaS products utilizing AI technology

Services

  • ZEROCK
  • TRAFEED
  • TIMEWELL BASE
  • WARP
  • WARP 1Day
  • WARP NEXT Corporate
  • WARP BASIC
  • WARP ENTRE
  • Alumni Salon
  • WARP for Schools
  • Alumni Ventures
  • FDE
  • AI Consulting
  • EMC GLOBAL

Company

  • About Us
  • Team
  • Why TIMEWELL
  • News
  • Contact
  • Free Consultation

Content

  • Insights
  • Knowledge Base
  • Case Studies
  • Whitepapers
  • Events
  • Solutions
  • AI Readiness Check
  • ROI Calculator

Legal

  • Privacy Policy
  • Manual Creator Extension
  • WARP Terms of Service
  • WARP NEXT School Rules
  • Legal Notice
  • Security
  • Anti-Social Policy
  • ZEROCK Terms of Service
  • TIMEWELL BASE Terms of Service

Newsletter

Get the latest AI and DX insights delivered weekly

Your email will only be used for newsletter delivery.

© 2026 株式会社TIMEWELL All rights reserved.

Contact Us
HomeColumnsテックトレンドClaude Code Superpowers Complete Guide to the Multi-Harness Development Methodology
テックトレンド

Claude Code Superpowers Complete Guide to the Multi-Harness Development Methodology

Published2026-02-22Updated2026-11-06Ryuta Hamamoto
AIClaude CodeDeveloper ToolsAI DevelopmentWARP

A look at where obra/superpowers stands as of 6 November 2026, based on the official repository and release notes. Covers the move from v6 to v7, supported harnesses and install commands, the 15 skills, how to pair it with the current Claude models, and the pitfalls of enterprise rollout.

Claude Code Superpowers Complete Guide to the Multi-Harness Development Methodology
Share

Hello, this is Hamamoto from TIMEWELL.

The short version: Superpowers is no longer a handy Claude Code plugin. It is a development methodology that lays the same discipline across several AI coding tools. The adoption question is not "do we install it" but "do we bend our development flow to this pattern." It also moves fast, and numbers and specs turn over within weeks. This article is based on the official repository and release notes as of 6 November 2026, so check the current state before you roll anything out. It is not a recommendation to adopt any particular tool.

When I wrote the first version more than six months ago, Superpowers was "one person's open-source plugin for bringing discipline back to Claude Code." Open the repository today and the description reads: "a complete software development methodology for coding agents."1 Not a plugin, but a methodology. That single change of wording captures what has happened over the past nine months.

As of 6 November 2026, GitHub stars stand at roughly 313,000 (about 257,000 at the July update)1. And it no longer targets only Claude Code. The same skills now run across more than a dozen harnesses, including Cursor, Antigravity, Codex, and GitHub Copilot CLI. From an individual's handy tool to a development discipline that an organization can standardize. This rewrite verifies where it stands now against primary sources. If you want to first gauge whether your organization is ready to take on this kind of standardization, the free AI readiness check is a good place to start.

What Superpowers Is (From Plugin to "Development Methodology")

Superpowers is MIT-licensed open source published under the GitHub handle obra (Jesse Vincent)1. In the first version I described it as an individual's plugin, but it is now a public asset maintained continuously, with releases several times a month. That distinction looks minor but matters: it feeds directly into the continuity and reliability you need before folding it into a business. I did not re-verify the author or the team behind it in this update.

The essence of the contents has not changed. It is a collection of "instruction books for the AI" written up as Markdown SKILL.md files. A meta-skill is injected at the start of a session so that Claude always consults a skill before acting. With test-driven development (TDD — the practice of writing a failing test before you implement) and systematic debugging at its core, it walks the AI through requirements refinement, design, planning, and implementation in sequence. The fact that it is designed as a set of composable skills is why Superpowers works as a "pattern."

The rename from "plugin" to "methodology" is, I feel, not merely wordplay. A plugin is a question of whether you install it; a methodology is a question of how a team works. To the field's familiar complaint that letting the AI write freely produces unstable quality, Superpowers responds not by adding a tool but by imposing discipline on the way development itself proceeds. That is exactly why the adoption debate shifts from "it's convenient, so let's add it" to the heavier decision of "do we bend our development flow to this pattern." I actually see that added weight as a change that is easier, not harder, for enterprises to handle, because the object of the decision moves from tool selection to development-process design.

Struggling with AI adoption?

We have prepared materials covering ZEROCK case studies and implementation methods.

Book a Free ConsultationDownload Resources

"Then vs. Now": What Changed Over Nine Months

Between February 2026, when I wrote the first version, and now, the premises have almost entirely turned over. First, scope. Back then it was effectively Claude Code-only; now multi-harness support is its biggest selling point. Next, version. It was v5 or earlier at first, v6.1.1 at the July update, and today the latest is v7.0.2, released 3 November 20262. GitHub stars have moved an order of magnitude too, from around 57,000 then to roughly 313,000 now1.

The model premises changed as well. The first version assumed Opus 4.7 and Sonnet 4.7, and the July update assumed Opus 4.8 and Sonnet 5. As of November 2026, Anthropic's current models are Fable 5.1, Opus 5.5, Sonnet 5.5, and Haiku 5.5. Opus 4.8 and Sonnet 5 are now listed as legacy models (still available)3. More on that in the model design section below.

I need to correct something I wrote in July. I said Gemini CLI "was fully dropped from v6.0.0 onward following Google's end-of-life." Checking the release notes, it was removed in v6.1.0 (June 30), in response to Google ending that CLI on June 18. Then v6.2.0 (July 24) restored Gemini CLI support, with the notes calling the removal "premature"2. The current README lists a Gemini CLI install command again1.

One more. The July version also said the bootstrap diagrams were replaced with prose in v6.0.0. I could not confirm that in the v6.0.0 release notes, so I am not asserting it. Numbers and dates get verified against the originator's primary source. Drop that principle and the ground gives way.

From v6 to v7: How the Releases Unfolded

The turning point was v6.0.0 on June 16, 2026. The notes for this full rewrite of the repository say execution is "roughly twice as fast" while spending "almost 50% fewer tokens." The same notes add that these numbers will not hold on every harness and every workload2. So measure in your own environment before you rely on them. The review stage also changed: the two reviewer prompts (spec-reviewer-prompt.md and code-quality-reviewer-prompt.md) were merged into a single task-reviewer-prompt.md.

Here are the main steps since. v6.0.2 (June 17) stopped shipping the evals submodule that had been breaking installs. v6.0.3 (June 18) moved the subagent-development scratch files, since Claude Code blocks agent writes to .git/. v6.1.0 (June 30) trimmed the bootstrap to improve token efficiency and added a Codex marketplace manifest. v6.2.0 (July 24) restored Gemini CLI support, and the finishing-branch menu stopped offering to discard work by default.

From August on, the story is harness expansion. v6.3.0 (August 12) added Devin CLI, Hermes Agent, and Grok Build CLI. v6.4.1 (September 19) introduced the diagnosing-superpowers skill, rebuilt executing-plans to run inline in one context, and added OpenCode 2.0, Muse, and Qwen Code. v6.4.2 (September 25) made writing-plans produce leaner plans that record decisions instead of full code, and the repository's own CLAUDE.md was replaced by AGENTS.md2.

The newest line is v7. v7.0.0 (October 10) rebuilt brainstorming to draw out what you want, play it back, and size the work, and added a native plugin manifest for Antigravity. v7.0.1 (October 21) adjusted it to write a design doc before building, and trimmed three long skills so they survive compaction on Claude Code. The latest, v7.0.2 (November 3), asks fewer questions in brainstorming and starts with why, offering options with a recommendation2. Along the version axis, v6 is "faster, cheaper, across multiple environments," and v7 shifts the weight to the quality of the first conversation.

Multi-Harness Support and the Correct Install Commands

Multi-harness support is still the headline feature. The README gives install steps for Claude Code, Antigravity, Codex (app and CLI), Cursor, Devin CLI, Factory Droid, Gemini CLI, GitHub Copilot CLI, Grok Build CLI, Kimi Code, OpenCode, Pi, Qwen Code, Hermes Agent, and Muse1. You can extend the same development discipline horizontally regardless of which tool a team uses. For anyone trying to lay down an organizational standard, that lands harder than you might expect.

The install command has changed from the old version, so get this right. The steps I wrote in the old version — /plugin marketplace add anthropic followed by /plugin install superpowers — are no longer correct. The current official marketplace name is claude-plugins-official. Here are the current commands for the main harnesses, copied from the README.

Harness Install command
Claude Code /plugin install superpowers@claude-plugins-official
Cursor /add-plugin superpowers
Antigravity agy plugin install https://github.com/obra/superpowers
Gemini CLI gemini extensions install https://github.com/obra/superpowers
GitHub Copilot CLI copilot plugin marketplace add obra/superpowers-marketplace, then copilot plugin install superpowers@superpowers-marketplace
Grok Build CLI grok plugin install superpowers@xai-official --trust
Kimi Code /plugins install https://github.com/obra/superpowers
Pi pi install git:github.com/obra/superpowers

Codex installs from the sidebar plugin screen or /plugins, and OpenCode works by having it read a dedicated INSTALL.md. When distributing internally, the quickest move is to paste this table into your onboarding materials, but commands can change with any release, so check the README before you distribute. Even on teams with a mix of harnesses, the skill implementation is a single thing, so you can unify review criteria and operating rules.

Where installed skills live inside the .claude folder, and how they divide labor with CLAUDE.md and hooks, is laid out in the complete guide to the .claude folder. If you want to dig into how skill definitions themselves are written, or into designing your own in-house skills, the convention I laid out in the guide to choosing Claude Code skills — "put only trigger conditions in the description and keep the body focused on procedure" — applies directly here.

Also readMastering Claude Code Sub-agents | Practical Patterns for Parallel Task Dispatch and Multiplied Coding Speed [2026 Edition]

Where the 15 Skills Fit and Model Design

The README currently lists 15 skills. In July I wrote 13; diagnosing-superpowers was added in September1. Add a one-line note on how each is used in practice and it looks like this.

Category Skill Where it fits in practice
Testing test-driven-development The core discipline: write a failing test first, then make the implementation follow the test
Debugging systematic-debugging A four-phase root-cause process, preventing guesswork fixes
Debugging verification-before-completion Force the AI to actually run and confirm before saying "done," killing false completion reports
Debugging diagnosing-superpowers Investigate what went wrong in a session, with evidence
Collaboration brainstorming Draw out intent before building. Small changes are agreed in chat; larger projects get a design document
Collaboration writing-plans Break an approved design into a plan of small tasks
Collaboration executing-plans Run the plan inline in one context, with one final review (the cheapest option)
Collaboration dispatching-parallel-agents Distribute independent tasks across parallel sub-agents
Collaboration requesting-code-review After implementation or before a merge, have it confirm spec compliance on its own
Collaboration receiving-code-review Verify the technical validity of review feedback before applying it, rather than swallowing it whole
Collaboration using-git-worktrees Isolate work from the current branch to prevent parallel-work accidents
Collaboration finishing-a-development-branch Wrap up merges, PRs, and cleanup after implementation, by the book
Collaboration subagent-driven-development One sub-agent per task, with two-stage review for spec compliance and then code quality (the most thorough option)
Meta writing-skills The foundation for creating new skills, including a testing methodology
Meta using-superpowers The entry point to the skills system, invoking existing skills correctly

The README's basic flow runs brainstorming, using-git-worktrees, writing-plans, then subagent-driven-development or executing-plans, then test-driven-development, requesting-code-review, and finally finishing-a-development-branch1. Choosing between executing-plans and subagent-driven-development is a choice between cheap and thorough. For how to think about designing with sub-agents, the practical guide to subagents is a useful companion.

Skills are the vessel for discipline; the model is what actually writes the code. So decoupling model design from the picture is a mistake. As of November 2026, Anthropic's current models are these four3.

Model Per million tokens (input / output) Anthropic's positioning
Claude Fable 5.1 USD 10 / USD 50 Demanding reasoning and long-horizon agentic work
Claude Opus 5.5 USD 4 / USD 20 Long-running agentic coding and knowledge work; the starting point for most workloads
Claude Sonnet 5.5 USD 2 / USD 10 The balance of speed and intelligence
Claude Haiku 5.5 from USD 0.10 / from USD 0.50 High-volume, latency-sensitive work such as classification, extraction, and routing

Sonnet 5.5 was released on September 28, 2026, Opus 5.5 on September 22, and Haiku 5.5 on October 745. In its announcement Anthropic describes Sonnet 5.5 as suited to well-scoped everyday tasks and Opus 5.5 as the choice for complex, open-ended work that needs careful judgment4. For the kind of Superpowers iteration that reads and rewrites the same files over and over, such as TDD and staged review, that split translates directly into cost. My own view is that routing light fixes and routine implementation to the lighter model, and design decisions and review to the higher tier, is an easy line to set as team policy.

Treat security work separately. Anthropic says that with Sonnet 5.5, higher-risk cybersecurity tasks will visibly fall back to Sonnet 54. When you handle vulnerabilities, check which model is actually answering. For vulnerability detection in AI-generated code itself, the cleaner framing is to hand that off to Anthropic's security product; I have summarized the operating patterns in the Anthropic Claude Security public beta implementation guide. Credentials and environment variables should not be entrusted to Superpowers but protected via the harness's own hooks and environment isolation — for that, see the Claude Code environment variables and security complete guide. If you want to lay the same discipline over company-wide knowledge work rather than just the engineering team, an enterprise AI platform like ZEROCK — with domestic AWS hosting, audit logs, and knowledge control — is one option worth considering.

Enterprise Adoption Pitfalls and Running an Organizational Standard with WARP

Superpowers is powerful, but distribute it company-wide as-is and problems are guaranteed. Across WARP-based engagements I have seen four pitfalls repeatedly. First, permission bloat. Some skills assume bash execution, worktree operations, and file deletion. Roll them out without checking against your IT policy and unintended operations can run against production repositories. It is safer to separate the skills you allow from those you forbid per business unit, and to bind permission scopes from the harness-side configuration. For concrete ways to bind them, the complete guide to Claude Code hooks is a good starting point.

Second, TDD enforcement becoming a hollow formality. The test-driven-development skill carries an iron rule: "you must not write implementation code without first writing a failing test." Teams that prioritize speed are tempted to downgrade this to a guideline, but doing so reverts to AI-driven implementation and quality degrades. Enforcing it or explicitly removing it — that binary choice causes less confusion for the team. Third, skipping the human approval gates. The requirements-refinement and design phases are, by design, gates for humans to agree on direction. Leaving those to the AI's internal reviewers alone usually means implementation proceeds on a wrong premise, producing large rework at the verification stage. At a minimum, keep a senior engineer's approval authority on the design and verify phases.

Fourth, managing the cost of review round-trips and sub-agents. subagent-driven-development spins up a sub-agent per task, so defaulting to the higher-tier model makes inference cost run hotter than expected. You need to set a cap on task size and route light fixes to Sonnet 5.5 — in other words, agree on a per-model cost policy up front. I dug into this point in the Claude Code agent teams operations guide as well. If you want to port Superpowers' sub-agent structure into your own workflows, the Claude Agent SDK implementation guide is also a useful starting point.

Making sure this design is not just "installed and distributed" is where WARP comes in. In TIMEWELL's AI implementation consulting service WARP, assuming the current Superpowers line (v7), we build an "organizational-standard operating design" together with you, fitted to your repository structure and IT requirements. Concretely: auditing existing repositories and development flows (permissions, TDD enforcement, review structure); distributing via the official marketplace and preparing internal guidelines; per-model cost and responsibility design across Opus 5.5 and Sonnet 5.5; embedding human approval into design and verification; and half-day to full-day workshops for engineering teams. Lately we hear especially often from companies that have thrown Superpowers in-house but are vague on who owns it and how far automation should go. Let us spend a 30-minute online session hearing your situation and mapping the next step. Reach out through a consultation to get started.

Unlike the first version, when I was chasing Superpowers as a tool, what I am now facing is the question of "how do you root a development pattern into an organization." Precisely because it has matured as a methodology, the skill of adoption is now what makes the difference. Before you rush to company-wide deployment, first run the pattern all the way through with a single team. That, I believe, is the fastest road even though it looks like the long way around.


TIMEWELL Co., Ltd. — Ryuta Hamamoto

Footnotes

  1. obra/superpowers, GitHub repository (description, skill list, install commands, star count, license; checked 2026-11-06). https://github.com/obra/superpowers ↩ ↩2 ↩3 ↩4 ↩5 ↩6 ↩7 ↩8

  2. obra/superpowers, Releases (changelog from v6.0.0 to v7.0.2; the v6.0.0 notes were checked on page 2). https://github.com/obra/superpowers/releases ↩ ↩2 ↩3 ↩4 ↩5

  3. Anthropic, Claude Developer Platform, Models overview (current models, pricing, legacy models; checked 2026-11-06). https://platform.claude.com/docs/en/about-claude/models/overview ↩ ↩2

  4. Anthropic, "Introducing Claude Sonnet 5.5" (2026-09-28; pricing, positioning, comparison with Opus 5.5, handling of cybersecurity tasks). https://www.anthropic.com/news/claude-sonnet-5-5 ↩ ↩2 ↩3

  5. Anthropic, Newsroom (lists Opus 5.5 on 2026-09-22 and Haiku 5.5 on 2026-10-07; checked 2026-11-06). https://www.anthropic.com/news ↩

This article was produced with the help of AI. A human verified the primary sources and edited the text before publication.

Read next

Articles that answer the question you are likely to have next.

  1. 1Inside the Claude Code .claude Folder (Late-2026 Edition): Designing Your Project's Brain by HandHave you ever actually opened the .claude folder in Claude Code? The way you design this folder all but decides the quality of your collaboration with the AI.
  2. 2Mastering Claude Code Sub-agents | Practical Patterns for Parallel Task Dispatch and Multiplied Coding Speed [2026 Edition]Mastering Claude Code Sub-agents | Practical Patterns for Parallel Task Dispatch and Multiplied Coding Speed [2026 Edition].
  3. 3The Complete Guide to Claude Code Hooks | Implementation Techniques for Accelerating Development with Tool Event Automation (2026 Edition)The Complete Guide to Claude Code Hooks | Implementation Techniques for Accelerating Development with Tool Event Automation (2026 Edition).

Ready to optimize your workflows with AI?

Take our free 3-minute assessment to evaluate your AI readiness across strategy, data, and talent.

Book a Free Consultation
Book a Free Consultation45-minute online sessionDownload ResourcesProduct brochures & whitepapers

Share this article if you found it useful

Share

Newsletter

Get the latest AI and DX insights delivered weekly

Your email will only be used for newsletter delivery.

Related Knowledge Base

Enterprise AI GuideAI Adoption Roadmap

Solutions

Solve Knowledge Management ChallengesCentralize internal information and quickly access the knowledge you need
AI Adoption & DX SupportEnd-to-end support from strategy to adoption

Learn More About テックトレンド

Discover the features and case studies for テックトレンド.

Contact UsView ZEROCK Details

Related Articles

Automating Code Review with Claude Code: A Practical Guide to Superpowers and Hooks [2026 Edition]

Automating Code Review with Claude Code: A Practical Guide to Superpowers and Hooks [2026 Edition]

Ten PRs piled up going into the weekend; more PRs landing before the previous ones are reviewed. Every engineering team has lived this.

2026-04-24
Inside the Claude Code .claude Folder (Late-2026 Edition): Designing Your Project's Brain by Hand

Inside the Claude Code .claude Folder (Late-2026 Edition): Designing Your Project's Brain by Hand

Have you ever actually opened the .claude folder in Claude Code? The way you design this folder all but decides the quality of your collaboration with the AI.

2026-03-25
Claude Code vs Cursor vs Cline: A Deep Comparison of the Big Three AI Coding Tools — Features, Pricing, and How to Choose [2026 Edition]

Claude Code vs Cursor vs Cline: A Deep Comparison of the Big Three AI Coding Tools — Features, Pricing, and How to Choose [2026 Edition]

A thorough comparison of Anthropic Claude Code, Cursor, and Cline (formerly Claude Dev) based on features, pricing, and real developer feedback as of April 2026.

2026-04-24
When AI Tools Become the Infection Vector: How the TanStack "Mini Shai-Hulud" Attack Redefined Software Supply Chain Risk

When AI Tools Become the Infection Vector: How the TanStack "Mini Shai-Hulud" Attack Redefined Software Supply Chain Risk

On the evening of May 11, 2026, a shocking piece of news swept across engineering timelines worldwide.

2026-05-13
Claude Code Enterprise Onboarding: A Complete Guide to Security, Cost, and Operations Across Six Rollout Steps (2026 Edition)

Claude Code Enterprise Onboarding: A Complete Guide to Security, Cost, and Operations Across Six Rollout Steps (2026 Edition)

Claude Code Enterprise Onboarding: A Complete Guide to Security, Cost, and Operations Across Six Rollout Steps (2026 Edition).

2026-04-25
The 52 Best Claude Code Skills: How to Pick Them and Install Them Safely

The 52 Best Claude Code Skills: How to Pick Them and Install Them Safely

A verified list of 52 Claude Code skills in four groups: bundled with Claude Code, Anthropic's official repository, financial services, and community-built. Every source, license, and last update was checked on October 4, 2026, alongside a practical guide to choosing, installing, and vetting skills.

2026-02-16