GitHub Agentic Workflows goes GA: read-only agents, with every proposed change checked before it lands
Oct 28, GitHub made Agentic Workflows generally available: Markdown-defined agents running inside Actions, read-only by default, sandboxed, with safe-output validation and a threat-detection job before any change applies.
GitHub Agentic Workflows, the GitHub Next project that runs coding agents inside GitHub Actions, is now generally available on Free, Team, and Enterprise. You describe a task such as issue triage, CI failure analysis, docs updates, or test writing in a Markdown file, and it compiles to ordinary Actions YAML, so it inherits your existing runners and policies. Copilot is the default engine, or you can bring your own agent.
The interesting part is the safety layering. Agents run read-only by default, inside a sandboxed container behind a firewall. They don't write directly: their outputs go through a "safe outputs" validation step, and a separate threat-detection job scans every proposed change before it is applied. The GA release adds self-hosted runner support, a non-experimental gh aw forecast that projects monthly AI credit spend including the threat-detection cost, per-workflow model choice and routing, and stronger isolation in micro-VMs for cross-repository work.
Why it matters for code review
This is the scheduled, always-on agent moving from hobbyist harness to platform default. Once agents run on every issue and every failed build, nobody reviews each action by hand, so the review step has to live in the pipeline: a narrow set of permitted outputs, plus an independent check on each one before it touches the repo. Notice that the checking job has its own line in the budget forecast. Verification under sustained volume is a cost you plan for, not a free side effect.