GitHub Copilot Code Review Learns Your Team's Rules: Agent Skills and MCP Now Generally Available

Add a SKILL.md to .github/skills/ and Copilot code review enforces your team's coding standards on every PR; MCP connections let it pull context from your issue tracker and docs — both went GA for all Copilot plans on July 29.

GitHub announced general availability of two Copilot code review extensions on July 29:

Agent Skills. Add a SKILL.md file under .github/skills/<skill-name>/ and Copilot code review applies those instructions to every PR it reviews in that repository. Skills can encode security patterns, naming conventions, architectural rules, or any standard your team enforces inconsistently in manual review. Review comments attribute which skill generated them, so reviewers can trace a comment back to the rule that fired.

MCP in code review (read-only). Copilot can now connect to your MCP servers to pull context from third-party platforms — issue trackers, service catalogs, internal documentation — into the review. The GitHub and Playwright MCP servers activate by default. All MCP tool calls are restricted to read-only.

Available to all Copilot Pro, Pro+, Business, and Enterprise users.

Why it matters: This brings the same "repo-level instruction" model that Claude Code users use with CLAUDE.md to Copilot code review. Teams running both tools can now apply consistent, machine-enforced standards across both automated review workflows from a single file per skill.

Sources: Copilot code review: Agent skills and MCP now generally available — GitHub Changelog · About agent skills — GitHub Docs · Using Copilot code review — GitHub Docs