goose-skills: a GTM skill library for Claude Code, Cursor and Codex

The gap goose-skills fills between a coding agent and a growth team

A coding agent is good at reading a repository and bad at knowing what a competitor's Meta ad library looks like this week. goose-skills is a library of prompt-shaped workflows, called skills, that give the agent that second kind of knowledge. The README describes the target user plainly: teams doing ads, social media, content, competitive intelligence, SEO, lead generation and GTM who already work inside Claude Code, Cursor or Codex. The repository ships Python and a skills catalog rather than a standalone SaaS dashboard, so the interface is the chat window you already have open.

The Brand Growth collection is the clearest statement of intent. According to the README it is not a separate package or command; it is a curated path through the normal catalog for consumer and ecommerce brand work, organized into four stages: Research, Analyze, Create, and Learn and iterate. That last stage is the interesting one. It tells the agent to bring performance results back into the analysis skills and decide the next test, which is the loop a growth team runs manually today.

How a skill gets from the catalog to your agent

The mechanism is a catalog lookup plus a download, not a long-running service. The README states that when you ask your agent to use a skill, the agent searches the GooseWorks catalog, downloads the skill, and runs it. The npm package exposes a binary named goose-skills, and the repository carries a generated skills-index.json plus scripts/build-index.js that produces it. Each skill directory follows a metadata contract, and node scripts/validate-skills.js checks SKILL.md against skill.meta.json. That contract is what makes the catalog searchable rather than a folder of loose prompts.

The data path for social and ad-library research is the part worth understanding before you commit. ScrapeCreators powers structured public social and ad-library research behind several workflows, and the README says signed-in GooseWorks users reach it through a managed first-party proxy without a separate ScrapeCreators key. The user-facing skills are expected to return a brief, a shortlist or a recommendation rather than raw API output. So the architecture is: local agent, local skill files, remote data proxy, and a credit meter in between.

Installing goose-skills and running a first skill

The README offers an agent-driven install and a manual one. The manual path is a single command that targets every detected agent:

npx gooseworks install --all

After that, authentication opens a browser session. The README does not document what happens if the browser flow is unavailable, so on a headless machine you should expect to solve that yourself.

npx gooseworks login

Once signed in, the skills are addressed through a slash command. The README's own example for a first real run is onboarding, which collects company context for later growth work:

/gooseworks onboard me

The README says the agent finishes onboarding by asking what you want to do first. From there, skills are invoked by name in natural language, for example asking for the competitor-intel skill to research a company. Two supporting commands are documented for day-to-day use: npx gooseworks search "reddit scraping" queries the catalog, and npx gooseworks credits reports your balance. If you prefer to build the index yourself rather than install, the README gives a source path with node scripts/validate-skills.js, node scripts/build-index.js and node bin/goose-skills.js list.

The credit balance is the real constraint, and the README is thin on it

Every workflow that touches social or ad-library data passes through a managed proxy, and the README documents npx gooseworks credits as a way to check your balance without saying what a credit costs, what a skill consumes, or what happens when the balance hits zero. That is the most consequential omission in the documentation. A team evaluating this for recurring competitive tracking cannot size the monthly cost from the repository alone; the README points to the hosted product instead.

There is a second limitation in the same area. Because the proxy is first-party and tied to a signed-in account, you cannot swap in your own ScrapeCreators key for the managed workflows. The README frames that as a convenience, and for a solo operator it is. For a team with an existing scraping contract, or one that needs to run in an environment without browser-based sign-in, it removes the option to bring your own credentials.

A third boundary is scope. The catalog is aimed at growth and GTM tasks. If your problem is code review, test generation or refactoring, the catalog's categories (Ads, SEO, Lead generation, Outreach, Content, Research, Competitive intel, Monitoring, Social, Brand) do not cover it, and you are paying the install and sign-in overhead for nothing.

How goose-skills differs from wiring your own MCP scraping tools

The obvious alternative is to skip the library and connect a coding agent directly to scraping and analytics APIs through MCP servers or custom tool definitions. That approach gives you full control over credentials, rate limits and data retention, and it costs nothing beyond the API bills you already pay. The difference in approach is where the judgment lives. With a custom tool setup, the agent gets raw records and you write the prompt that turns them into a brief. goose-skills moves that step into the skill itself: the README says the user-facing skills turn source data into a brief, shortlist, analysis or recommendation rather than returning raw API output. You are trading configuration control for pre-written analysis, and the trade is only worth it if the pre-written analysis matches how your team actually reasons about a competitor's ad account.

Maintenance, upgrades and what the MIT licence does and does not cover

The last push to the repository was on 2026-09-01, and the repository is not archived. The package version in package.json is 1.0.1. Upgrades are handled by a documented command, npx gooseworks update, which the README describes as updating to the latest skill version. Because skills are downloaded at use time rather than vendored into your project, a catalog change can reach your agent without a commit in your repository. That is convenient and also means your prompt results can shift without a diff to review.

The licence is MIT, which permits commercial use, modification and redistribution with the copyright notice and permission notice preserved. That covers the skill files and scripts in the repository. It does not cover the hosted proxy, the credit system or the Gooseworks account you sign into, which are separate services governed by their own terms. The README does not describe a self-hosted path for the proxy, so you should treat the MIT grant as covering the client side only. Nothing here is legal advice; read the LICENSE file and the service terms yourself.

Editorial conclusion

Adopt goose-skills if your growth work already happens inside a coding agent and you want research, ad analysis and content tasks to run from the same prompt surface, accepting that the social data path goes through a signed-in GooseWorks proxy and a credit balance. Do not adopt it if you need self-hosted scraping with your own ScrapeCreators key, or if your team does not use Claude Code, Cursor or Codex at all. Before committing, run npx gooseworks install --all in a scratch directory, confirm the agent picks up the /gooseworks command, and check npx gooseworks credits to see what the balance looks like on a free account.

Frequently asked questions

Official sources

1. gooseworks-ai/goose-skills on GitHub
2. Issues
3. License: MIT
4. Project website
5. README