GitHub Issues adds agent automation controls (public preview)

GitHub announced a public preview of agent automation controls in GitHub Issues. As automations increasingly label, type, assign, and close issues for users, the new features surface the reasoning and let you review changes before they apply.

Key points

• Approvals: configure automations to suggest instead of apply; changes wait in a panel on the issue for you to accept or decline individually or all at once.
• Confidence: agents rate each action high/medium/low; high-confidence changes apply automatically while medium and low are held as suggestions.
• Rationale: every supported action records its reason, giving an audit trail of what changed and why.
• Controls: use has:suggestions in issue search to find pending changes; repo admins set the confidence threshold that governs auto-apply.
• Scope: works with GitHub Agentic Workflows and Copilot cloud agent, plus REST/GraphQL APIs; covers labels, fields, type, close, and assignees.

Note: GitHub stresses Approvals is a workflow convenience, not a security control - an agent with permission to change issues can apply changes directly. Details are in GitHub's changelog - see Read original.