Cursor for Teams: Setting Up Shared Skills via Git

An individual developer using Cursor gets faster. A team using Cursor with shared skills gets a compounding advantage that grows with every member. Here is how to set up team skills that turn institutional knowledge into automatic capability.

The biggest untapped opportunity in Cursor is not individual productivity; it is team productivity through shared skills. When a team encodes its conventions, patterns and knowledge as skills committed to git, every developer's AI assistant applies that knowledge automatically. The result is consistency, faster onboarding, and institutional knowledge that stops living only in senior developers' heads.

Why Team Skills Compound

Individual skills help one person. Team skills help everyone, and they get better over time. Every developer who encounters a gap (the AI did not apply a convention because no skill covered it) can add a skill that fixes it for the whole team. Over months, the skill library becomes increasingly comprehensive. A new hire on day one benefits from years of accumulated team knowledge, applied automatically by their AI assistant.

Step 1: Create a Team Skills Repository

The simplest pattern is a dedicated directory in your main codebase: .cursor/skills/ committed to git. For larger organisations, a separate private skills repository that multiple projects reference also works. The key is that the skills live in version control where the whole team can access, update and review them through normal git workflows.

Step 2: Document Team Patterns as Skills

Identify the knowledge worth encoding: code style conventions, internal API usage patterns, deployment procedures, on-call runbooks, domain-specific business logic. Each becomes a skill folder with a SKILL.md describing what it does and when to apply it. Start with the patterns that new team members most often get wrong, since those deliver the most immediate value.

Step 3: Onboarding Becomes Automatic

With skills in the repository, onboarding changes fundamentally. A new developer clones the repo and immediately has every team convention installed in their Cursor. Before they write a line of code, their AI assistant already knows how your team names things, structures APIs, handles errors and deploys. The weeks normally spent absorbing tribal knowledge compress dramatically because the knowledge is applied automatically rather than learned through trial and error.

Step 4: Evolve Skills Through Pull Requests

Skills evolve like any other code: through pull requests. When someone identifies a missing pattern or an improvement, they open a PR adding or updating a skill. Reviewers approve, everyone pulls, and the new knowledge propagates to the whole team. This keeps skills current and turns knowledge management into a normal part of the development workflow rather than a separate documentation chore nobody maintains.

The Long-Term Payoff

Teams that treat their skills repository as a first-class asset, maintained and improved like production code, gain a compounding advantage. Their AI assistance is consistently better than teams relying on default behaviour, their onboarding is faster, and their accumulated knowledge is resilient to individual departures because it lives in skills rather than in people's heads. Over a year or two, this difference becomes substantial.