wondelai/skills: 65 Agent Skills That Turn Bestselling Frameworks Into Claude Code and Codex Workflows

What wondelai/skills actually ships, and who it is for

The repository is a collection of 65 agent skills, described in the README as 51 expert frameworks from bestselling books and industry style guides plus 14 metaskills. The frameworks are familiar names: Jobs to Be Done, The Mom Test, Refactoring UI, Clean Code, Domain-Driven Design, Designing Data-Intensive Applications, Crossing the Chasm, Blue Ocean Strategy, Influence Psychology, Lean Startup, Release It! and others. Each one is a directory at the repository root, such as jobs-to-be-done/ or clean-architecture/.

The intended audience is people who already work inside a coding or product agent and want a book's method available as a callable unit rather than a prompt they retype. The README lists Claude, Claude Code, Claude Cowork, Codex, Cursor, OpenClaw, Hermes Agent and other agentskills.io-compatible agents as targets, and the repository carries configuration directories for several of them: .claude-plugin/, .cursor/, .agents/, .pi/ and .windsurf/.

The 14 metaskills are the more unusual half. Their names are verbs: create-business, create-website, create-app, improve-business, improve-website, improve-app, grow-business, grow-website, grow-app, improve-code-quality, remove-technical-debt, design-code-architecture, conversion-optimization and architecture-optimization. A metaskill is not a framework summary; it is a journey that orchestrates the framework skills step by step. That distinction is the reason to look at this repository rather than a folder of markdown notes.

How a metaskill orchestrates the framework skills

According to the README, a metaskill asks you the decision questions phase by phase and keeps its state in your project's docs/ folder, so a journey survives across sessions. That is the whole mechanism, and it is worth reading twice. The agent does not run a fixed pipeline. It walks you through phases, asks questions at each one, and writes progress into a directory inside your own repository.

The state-in-docs choice has consequences the README does not spell out. Because the state lives in your project rather than in the agent's session memory, a journey can be resumed in a new session, and the artifacts are reviewable in a pull request like any other file. The same property means the docs/ folder becomes part of your working tree: if you commit it, the journey is versioned; if you do not, it is local scratch. The README does not document how to reset or roll back a journey, so plan for that yourself.

The framework skills underneath are simpler. Each is a self-contained directory, and the repository's top-level listing shows the same layout repeated dozens of times. The metaskills are the layer that decides which of them to invoke and in what order.

Installing via the Claude Code plugin marketplace

The first documented install path is the Claude Code plugin marketplace. You add the marketplace once, then install collections. Each collection bundles related frameworks: product-strategy carries Jobs to Be Done, Negotiation and The Mom Test; code-craftsmanship carries Clean Code, Refactoring Patterns, Software Design Philosophy, Pragmatic Programmer, DDD, Working with Legacy Code and Google Docs Style.

/plugin marketplace add wondelai/skills
/plugin install product-strategy@wondelai-skills
/plugin install code-craftsmanship@wondelai-skills

After the first command the marketplace is registered under the name wondelai-skills. The install commands then pull the named collection. The README lists ten collections in total, including ux-design, marketing-cro, sales-influence, product-innovation, strategy-growth, team-motivation, systems-architecture and metaskills. Installing metaskills@wondelai-skills brings in the 14 guided journeys rather than the individual frameworks.

If you only want one framework, a collection is more than you asked for. The plugin path is built around bundles, so the per-skill route below is the finer-grained option.

Installing individual skills with npx skills add

The second documented path is skills.sh, driven by npx. The README gives two forms. The --all flag installs everything, and the per-skill form installs one directory by name.

npx skills add wondelai/skills --all --global
npx skills add wondelai/skills/jobs-to-be-done --global
npx skills add wondelai/skills/refactoring-ui --global

The first command installs all skills; the second and third install single frameworks. The --global flag appears on every example in the README, which suggests the default target is a global skills location rather than the current project, though the README does not describe a project-local alternative. The same npx form covers the metaskills, for example npx skills add wondelai/skills/improve-code-quality --global.

Codex users are pointed at the same commands, because the README states that Codex adopted the same open SKILL.md standard and that Codex is a supported agentskills.io agent. The repository also ships a Codex plugin marketplace at .agents/plugins/marketplace.json with the same ten collections, which the README says Codex auto-discovers in a clone.

Where the skill set stops being the right tool

The clearest boundary is agent compatibility. The skills are SKILL.md files for agentskills.io-compatible agents. If your assistant is not one of those, this repository gives you prose you would have to adapt by hand, and the install commands have nothing to attach to. The README names Claude, Claude Code, Claude Cowork, Codex, Cursor, OpenClaw and Hermes Agent; anything outside that list is unverified.

A second boundary is the nature of the content. These are frameworks from books and style guides, not executable tooling. A skill can tell an agent to apply Jobs to Be Done or the Google developer docs style guide, but it cannot lint your repository or fail a build. If you want automated enforcement, you want a linter or a test suite, not a skill.

A third is the metaskill interaction model. Because a metaskill asks decision questions phase by phase, it is a poor fit for unattended runs. Handing create-business to an agent and walking away is not what the README describes. The state in docs/ also means a metaskill writes into your repository, so you should decide early whether that folder is committed or ignored. The README does not document rollback, so treat the first run as something to inspect rather than trust.

How this differs from writing your own SKILL.md files

The obvious alternative is authoring your own skills. The SKILL.md format is open and the README notes that Codex adopted the same standard, so nothing stops you from writing a directory that encodes your team's review checklist or your internal API conventions.

The difference is scope and maintenance. A hand-written skill captures what your team already knows; wondelai/skills ships 51 frameworks that most teams do not have written down, plus a metaskill layer that sequences them. Writing your own equivalent of the metaskills is the expensive part, since it means encoding phase-by-phase questions and a state convention in docs/.

The trade-off runs the other way too. A self-written skill can encode your actual conventions, and this collection cannot. The frameworks are general by design, drawn from published books and style guides, so they will not know that your service uses a particular migration pattern or that your design system has its own spacing scale. The practical split is to install the framework skills you recognize and write your own for anything specific to your codebase. The repository is MIT-licensed, so reading a skill directory to see how one is structured is a legitimate way to start.

Maintenance, licence and what a version bump means

The repository is not archived, and the last push was on 2026-08-29. The release history is recent and granular: v1.8.0 added technical-documentation following the Google developer docs style guide on 2026-08-29, v1.7.0 added conversion-optimization and architecture-optimization on 2026-08-28, and v1.6.0 added Agent Plugins support on 2026-08-10.

Those version notes tell you what an upgrade actually changes. Adding a skill is additive: a new directory appears and existing skills are untouched, so updating is low risk. The v1.6.0 entry is the different kind, since it concerns packaging against the Agent Plugins specification rather than content. If you installed through the Claude Code plugin marketplace or through npx skills add, your upgrade path is re-running the same install command; the README does not describe a version pinning mechanism, so you get whatever the current state of the repository provides.

The licence is MIT, stated in the repository and present as a LICENSE file at the top level. That is permissive and standard for this kind of content, but the frameworks themselves are derived from published books and style guides. The README attributes them by name. Whether a given skill's treatment of a copyrighted method is acceptable for your use is a question for your own review; the repository's MIT licence covers the repository's files, and this article is not legal advice.

Editorial conclusion

Adopt wondelai/skills if you already run Claude Code, Codex or Cursor and want named frameworks such as Jobs to Be Done or Clean Code available as invocable skills instead of pasted prompts. Skip it if your agent is not agentskills.io-compatible, or if you expect the metaskills to run unattended: they ask decision questions phase by phase and write state into your project's docs/ folder. Before installing, open the individual skill directory and read its SKILL.md to confirm the framework matches your problem, and check the LICENSE file for the MIT terms.

Frequently asked questions

Official sources

1. License: MIT
2. Project website
3. README
4. Releases
5. wondelai/skills on GitHub