Your team's expertise, built into every AI agent
AI coding assistants are fast, but they don't know how your team works. Skills give them your standards, decisions and hard-won lessons — so the output fits your codebase the first time, in GitHub Copilot, Claude Code or any other tool.
A capable agent is still a generic one
Without context, agents fall back on the most common patterns online — outdated APIs, the wrong test library, architecture that doesn't match yours.
People paste the same instructions into chat after chat. It is slow, easy to forget, and different for every person.
Two developers ask the same question and get two different designs. Reviewers spend their time correcting the agent instead of the problem.
A skill is a playbook the agent loads when it needs it
A skill is one SKILL.md file: a short description of when to use it, and a senior-level playbook of decisions, rules, examples and the expected output format.
- 1Agent sees a one-line description
At the start of a session the agent only reads each skill's name and description — a few lines each.
- 2Your request matches a skill
"Write tests for this form" matches
react-testing, so the agent loads the full playbook on demand. - 3It works the way your team does
The agent follows the skill's decisions and anti-patterns and returns output in the agreed format.
What teams get
Every agent follows the conventions, patterns and review checklists your team agreed on — not whatever the model saw most on the internet.
Stop pasting the same "we use Vitest, not Jest; always…" paragraph into every chat. The skill carries that context, so a short request is enough.
Skills encode decisions, trade-offs and anti-patterns. Output starts closer to what a senior reviewer would accept, so there are fewer review rounds.
New joiners get the team's judgement built into their tools from day one — and can read the same SKILL.md to learn why things are done that way.
One open format for GitHub Copilot, Claude Code, Codex, Cursor, Gemini CLI and more. Teams keep their tool; the knowledge is shared.
Agents only see a short description until a skill is relevant, then load the full playbook. Dozens of skills cost almost nothing when unused.
Skills live in Git with pull requests, code owners and CI validation. Guidance is versioned, auditable and easy to improve.
When someone solves a problem well, it becomes a skill everyone benefits from. Expertise stops living only in a few people's heads.
Same request, with and without a skill
Prompt: Write tests for the checkout form component.
Without
- Picks a test runner and mocking approach at random, often not the one in your repo
- Tests implementation details, so tests break on every refactor
- Mocks
fetchby hand in each test - Misses loading, error and accessibility states
- You explain your conventions, then ask again
With react-testing
- Uses the team's stack (Vitest, Testing Library, MSW) and file layout
- Tests behaviour the way a user sees it, with accessible queries
- Network mocked once at the boundary, reused across tests
- Covers happy path, validation, errors and loading
- Follows the agreed structure — ready for review
Illustrative example based on what the react-testing skill instructs. Results depend on the agent and the codebase.
How to show it's working
Measure a baseline before rollout, then compare after a few sprints. These are the signals that move first — track them per team, not per person.
| Measure | Why it matters |
|---|---|
| Review rounds per pull request | Fewer "please follow our convention" comments means skills are being applied. |
| Time from first commit to merge | Better first drafts shorten the loop between author and reviewer. |
| Time to first merged PR for new joiners | Onboarding speed is where encoded team knowledge shows up first. |
| Share of teams with skills installed | Adoption comes before impact — track it per team. |
Useful at every level of the team
Get code that fits the codebase with shorter prompts, and spend less time fixing agent output.
Encode standards once instead of repeating them in every review.
Work productively in an unfamiliar stack with the team's judgement built into their tools.
More consistent quality across teams and tools, and practices that survive people moving on.
Install the skills for your stack in one command
npx skills add AGCO-Global/org-skills