How Codenskills works

Codenskills is governed skill infrastructure for company Workspaces. Your company brings complete skill packages into one controlled place, organizes and reviews them, approves trusted versions, and lets connected AI clients install exactly what policy and grants allow.

The model in one minute

• A skill is a complete package with a required SKILL.md, imported from an upload, a GitHub repository or a connected MCP client.
• Every change produces a version that moves through Draft → Review → Ready → Superseded. Ready versions are the governed distribution target; an author can privately test only their own current Unreviewed Draft before submission.
• Skills live in one company Workspace with named users, roles, reviews, collections, security policy and access grants.
• Clients connect through MCP. Direct clients such as Claude Code, Codex, Cursor or VS Code authenticate with a workspace-bound access key. Hosted ChatGPT and Claude surfaces use OAuth.
• Authentication does not create a Workspace. Codenskills provisions the company Workspace and invites its first Owner; additional members join by invitation or configured company SSO.

Where to go next

• Quickstart — from company invitation to a connected client.
• Connect agents — access keys, MCP configuration and hosted connectors.
• Skills & versions — imports, versions and reviews.
• Skill security — package validation, trusted scans and release gates.
• Collections & sources — curation and provenance.
• Workspace governance — members, roles, groups and access control.
• Access & commercial terms — provisioning, licensed capacity and agreements.
• Security policies — organization criteria and enforcement.
• Notifications — in-app and email governance events.
• Feedback & support — report a problem or send an idea.
• Access keys — scopes, expiry and rotation.
• Enterprise SSO — company identity-provider access.

For the security model, see the security page. For company access, email notifications@codenskills.com.