Build a Team Skills Repository — Institutional Knowledge That Compounds

The most valuable skills pattern for organisations is not public publishing — it is the private team repository. Every team member contributes patterns. Every new hire benefits from accumulated knowledge from day one. Here is how to set it up.

Public skill marketplaces like skills.sh are excellent for general patterns — React best practices, document creation, common workflows. But the highest-value skills in any organisation are the ones specific to your codebase, your team conventions, your domain. These should not be public. They should be team assets.

The pattern is straightforward: a single private GitHub repository for team skills. Onboarding includes git clone of the repo and copying the skill folders to each team member AI agent skills directory. Updates propagate via git pull. The simplicity is the entire point.

Why This Pattern Compounds Over Time

Traditional documentation — wikis, internal blogs, onboarding documents — sits in places where humans read it parallel to their work. People reach for it when they remember to, often after they have already made the mistake the documentation would have prevented.

Skills sit where AI agents apply them directly. The same content becomes more impactful because it is integrated into the workflow rather than alongside it. When a developer asks their AI to write a function, the team code-style skill applies automatically. When they ask for a PR description, the team PR conventions are followed without explicit prompting.

Every senior team member who builds a useful skill adds to the shared collection. Every new hire benefits from accumulated organisational knowledge from day one. The knowledge stops being tribal — held in the heads of senior people — and becomes executable and shared.

What Goes in a Team Skills Repo

The skills that benefit most from being team assets fall into four categories.

Code style and conventions. Your team specific patterns for naming, error handling, logging, comments, file organisation. The agent applies these automatically every time it generates code.

Internal API patterns. How your team makes HTTP requests, handles authentication, structures error responses, designs endpoints. Critical knowledge that is currently in senior developers heads.

Operational runbooks. Deployment procedures, on-call response patterns, incident triage. Skills that make new on-call engineers as effective as senior ones from day one.

Domain knowledge. Your specific industry conventions, your customer types, your business logic patterns. The institutional knowledge that takes years to develop and is the hardest to transfer.

The Onboarding Workflow

For new hires, the skills repository becomes part of the standard onboarding checklist. After setting up their development environment, they clone the team skills repo and copy the folders to their AI agent skills directories. Day one, before they have written any code, the AI agent they use already knows the team conventions and applies them automatically.

Updates flow through the same process as code: developers open pull requests when they want to add or modify a team skill. Reviewers approve. Everyone pulls. New patterns become available to the whole team immediately. The knowledge is alive — it grows as the team grows.

What to Track

The metric to watch is hit rate — how often skills trigger when they should. Junior developers often discover gaps where the AI did not apply team patterns because no skill covered that situation. Each gap becomes a new skill or a description refinement. Over months, the team library becomes increasingly comprehensive and increasingly valuable.