Agent Upstart

Cool things people are building with agents — and practical help getting the most out of them.

Playbooks · Coding tasks

GitHub Agentic Workflows are now GA

Written November 7, 2026

Agentic Workflows went generally available on October 28. Build a self-maintaining test repo with real logs and permission boundaries — and mind the limits.

What went generally available

GitHub Agentic Workflows, an open-source project from GitHub Next, went generally available on October 28, 2026. A workflow automates reasoning-based tasks — issue triage, CI failure analysis, documentation updates, test writing — by running coding agents inside GitHub Actions. You define the automation in natural-language Markdown files, and the tool compiles them into standard Actions YAML. Because they are GitHub Actions, they reuse your existing runner groups and policy constraints. You can use GitHub Copilot or bring your own coding agent to power each workflow.

Availability is on github.com for GitHub Free, Team, and Enterprise. Using the default GitHub Copilot CLI engine to power a workflow requires a Copilot licence. Third-party engines — Claude, Codex, Gemini, Pi — are usable without one.

The safeguards, as GitHub describes them

Agents access GitHub content respecting integrity-filter rules, run with read-only permissions by default, and execute inside a sandboxed container behind an Agent Workflow Firewall. Outputs are validated through a safe-outputs process, and a dedicated threat-detection job scans all proposed changes before they are applied. That is the permission boundary to state up front when you set a workflow up: read-only unless you widen it, sandboxed, and with proposed changes scanned before they land.

Since the public preview, GitHub highlights several changes. Actions Runner Controller and Docker-in-Docker support now cover self-hosted ARC runners. The forecast command (gh aw forecast) is no longer experimental, and now includes both threat-detection credits and monthly ranges for a fuller picture of AI credit spend. The model inventory is refreshed, with model providers picked per workflow, and new workflow templates can be added with gh aw add-wizard. Model routing optimizes for task type and difficulty. And agent isolation is improved: agents can run in a Cloud Hypervisor or NVX micro-VM, with enclaves preventing data leaks in cross-repository workflows.

Build a self-maintaining test repo

The way to learn the tool is one workflow on a throwaway repo, not a production codebase. Pick one real test gap — a module with no tests, a flaky test nobody owns — and write one workflow for it in plain language: what the agent should do when a test fails, what it may change, and what it must leave alone. Keep the default read-only permissions for the first run, and read the logs and the threat-detection scan before letting any proposed change through.

GitHub's own guidance points to the quickstart guide to install the CLI extension and trigger a first workflow in minutes, and to the prebuilt workflows in GitHub Next's agentics repository for triage, dependency management, code review, and reporting. Start from a prebuilt workflow and narrow it, rather than writing one from scratch.

GitHub quotes an Accenture manager saying one lead saved an estimated 29 hours a week on development work, velocity more than doubled across his team, AI authors 91% of the team's commits, and production errors fell from about 11,000 a day to under 1,000. That is GitHub quoting a customer, not an independent assessment, and it describes one team, not a typical result.

Limits

Related playbooks

Sources

Back to: Coding playbooks hub · Also: OpenBot: GitHub events as workflows · pull_request_target: December enforcement · OpenAI's Agents API: the harness is now hosted