GitHub Copilot CLI now supports custom agents defined in Markdown files to automate repeated development workflows.

GitHub Copilot CLI is adding custom agents—AI assistants configured via Markdown files stored in a repository's `.github/agents` directory. Instead of one-off prompts, teams can encode their standards, tools, and guardrails into reusable agent profiles that behave consistently across the CLI, IDE, and GitHub.

Custom agents are defined using agent profiles (Markdown files with YAML frontmatter) that specify the agent's role, which tools it can access, and what guardrails apply. Examples include a security audit agent that runs standard security checks and produces a pull-request-ready checklist, or an infrastructure-as-code compliance agent that reviews Terraform plans and Kubernetes manifests against organizational policies.

Because agent profiles live in the repository, teams can review, version, and share them so the same expectations follow workflows from the terminal into pull requests. The agent will execute the workflow the same way every time it runs.