Daily brief
GitHub Lets Agents Wait for Pull Request Events, Copilot Plugins Go Team Wide, and Copilot Evals Opens in Preview
By Steven Sorensen ยท 19 Nov 2026

At a glance
- Agents and scripts can now wait for a pull request review or check result from GitHub CLI without asking GitHub over and over.
- Copilot Business and Enterprise teams can pick shared plugins once, and Copilot can connect to eight more services such as Supabase and Buildkite.
- Copilot Evals is in public preview, so you can rerun the same test after you change a skill, agent, or tool.
- Gemini CLI reached a new latest release, Codex stays on its current stable release, and Copilot CLI is unchanged.
GitHub posted the CLI news at about 7:00 PM ET Wed Nov 18, team customizations and the new connectors landed earlier that day, and Copilot Evals opened in public preview at about 1:00 PM ET Tue Nov 17.
Treat today as a stop polling, start sharing day: join the Event Streams waitlist if your agents babysit pull requests, choose a small set of team plugins before members each build their own, and try Copilot Evals on one skill or tool you already rely on.
Top Stories
1. GitHub CLI can wait for pull request events instead of polling
GitHub CLI v2.105.0 adds a gh event command set built on GitHub Event Streams, which opened as a preview on Tuesday. You run gh event listen with a pull request URL and the events you care about, such as a submitted review, and it creates the subscription, opens the stream and prints each event as a line of JSON. By default it exits after the first event, so it drops into a script or an agent loop, while --count, --follow and --timeout cover longer waits.
Supported events are submitted reviews, pull request state and head changes, merge conflicts and completed check suites. GitHub says gh pr view, gh run watch and related commands now suggest gh event listen to agents instead of polling. Event Streams needs no public webhook endpoint, so an agent on your laptop can receive events, but access goes through a waitlist and the commands are in preview and may change.
Practical dev impact: If you have an agent or script that checks a pull request every minute for a review or a green check, this is the replacement to plan for, because each poll spends rate limit for answers that are mostly unchanged. Upgrade gh, join the waitlist, and keep your polling loop until access is approved.
2. Copilot plugins can now be set once for a whole team
Team customizations are now generally available for organization teams on Copilot Business and Copilot Enterprise. Team maintainers and organization owners open the team on github.com, select the Customize Copilot tab, and choose approved Agent Plugins as team defaults, from built in, enterprise or team added marketplaces. Those plugins are then on by default for the team's direct members in VS Code, the GitHub Copilot app and Copilot CLI, though members of child teams do not inherit them and members can turn a team plugin off where policy permits.
Enterprise and organization policies still win. Later on Wednesday GitHub also added eight services that Copilot can connect to through MCP servers verified by GitHub, in public preview for all plans: Box, Buildkite, Calendly, ClickUp, Mixpanel, Supabase, Webflow and Wix. On Business and Enterprise an admin must enable the Editor preview features policy, and an MCP allowlist that names connectors one by one needs the new ones added.
Practical dev impact: Team leads can now hand everyone the same starting setup instead of a wiki page of install steps, so pick two or three plugins the team already trusts and stop there. Admins with a strict MCP allowlist should decide on the new connectors before developers ask why Supabase or Buildkite will not connect.
3. Copilot Evals is in public preview for testing your setup
GitHub Copilot Evals is now in public preview for Copilot Individual, Copilot Business and Copilot Enterprise. An eval is a repeatable test for AI behavior: you give an agent a realistic task, describe what a good result looks like, and check both the result and how the agent got there. Because output can change from one run to the next, GitHub says one good answer does not tell you whether a customization works well, so you rerun the same eval after you change a skill, agent or MCP tool to catch regressions.
You can ask Copilot in your own words or use the /eval command to create an eval for a skill, custom agent or MCP tool in your project, run it and see which checks passed and failed, find out why a check failed, and compare runs from before and after a change. You approve each run before it starts. GitHub Copilot Evals uses Vally, an extensible local eval tool, as its evaluation engine. You can install GitHub Copilot Evals as a Visual Studio Code extension, as a Copilot CLI plugin, or in the GitHub Copilot app by adding the github/evals marketplace from Customize.
Practical dev impact: Pick one skill, custom agent or MCP tool you already rely on and create an eval for it first. Each run waits for your approval, so start small. Rerun that same eval after you change the customization, so you can see whether the change helped or slipped.
Practical Impact Analysis
Our read: the GitHub CLI story leads because Event Streams and gh event listen give agents a way to wait for pull request events instead of polling, even though access still goes through a waitlist and the commands are in preview. Team customizations come second because they are generally available now for every Copilot Business and Enterprise organization. Copilot Evals is narrower, but it gives Individual, Business and Enterprise users a repeatable way to check a skill, custom agent or MCP tool after they change it.
The common thread is that today's changes give teams shared Copilot defaults and give agents clearer ways to wait on or test work, so it is worth knowing who owns each setting before an agent depends on it. Elsewhere, Claude Code v2.1.328 is the new latest build on npm, while the stable tag still points at v2.1.311. Gemini CLI v0.69.0 reached the latest tag Wednesday evening, and its v0.70.0 preview belongs on a test machine only. Codex stays on v0.170.0 as its stable release, so the v0.172.0 alpha builds are for test machines only, and Copilot CLI is unchanged at v1.0.114.
Tutorial
- Open a test shell. Every command below only reads versions or public package data.
- Script step 1 prints the npm tags to compare with our snapshot.
- Script step 2 shows whether you have GitHub CLI v2.105.0 or later.
- Script step 3 prints help and the event types for gh event listen, which works before you have Event Streams access.
- After waitlist approval, try the commented listen line with a real pull request URL.
# We have not run this script end to end on every platform. It prints and checks, and changes no files.
# Run it in a test shell. Nothing is written to your shell profile.
# 1. Snapshot at 6:03 AM ET Thu Nov 19 2026: claude-code latest 2.1.328, stable 2.1.311,
# copilot latest 1.0.114, codex latest 0.170.0, gemini-cli latest 0.69.0.
python3 -c 'import json, urllib.request; d = json.load(urllib.request.urlopen("https://registry.npmjs.org/-/package/@anthropic-ai/claude-code/dist-tags")); print("claude-code latest", d["latest"], "stable", d["stable"])'
python3 -c 'import json, urllib.request; d = json.load(urllib.request.urlopen("https://registry.npmjs.org/-/package/@github/copilot/dist-tags")); print("copilot latest", d["latest"])'
python3 -c 'import json, urllib.request; d = json.load(urllib.request.urlopen("https://registry.npmjs.org/-/package/@openai/codex/dist-tags")); print("codex latest", d["latest"])'
python3 -c 'import json, urllib.request; d = json.load(urllib.request.urlopen("https://registry.npmjs.org/-/package/@google/gemini-cli/dist-tags")); print("gemini-cli latest", d["latest"])'
# 2. GitHub CLI on this machine.
if command -v gh >/dev/null; then
gh --version
else
echo "no gh on PATH"
fi
# 3. The new event commands, readable before Event Streams access.
if command -v gh >/dev/null; then
gh event listen --help
fi
# After waitlist approval, wait up to 30 minutes for one submitted review:
# gh event listen --resource https://github.com/OWNER/REPO/pull/123 --event com.github.pull_request.review.submitted.v1 --timeout 30m
echo "Done. Nothing was installed or changed on disk."
If gh says the event command is unknown, you are on a version older than v2.105.0. For Copilot Evals, follow the install steps in GitHub's announcement for Visual Studio Code, Copilot CLI or the Copilot app.
Sources:
- Wait for pull request events with GitHub CLI (GitHub Changelog)
- GitHub CLI v2.105.0
- GitHub Event Streams is now in preview (GitHub Changelog)
- Customize GitHub Copilot for your team (GitHub Changelog)
- Connect Supabase, Box, Calendly, and more to Copilot (GitHub Changelog)
- GitHub Copilot Evals is now in public preview (GitHub Changelog)
- Claude Code changelog
- Gemini CLI v0.69.0
- Codex v0.170.0
Recommended AI prompt
Copy this paragraph into ChatGPT, Claude, Gemini, Grok, or whatever AI you use.
You are my developer tooling lead for Thu Nov 19, 2026. Story 1: GitHub CLI v2.105.0 adds gh event listen, which waits for pull request events such as a submitted review or a finished check suite through GitHub Event Streams, a preview behind a waitlist. Story 2: Copilot team customizations are generally available for Business and Enterprise, so team maintainers can set default plugins for direct members, and Copilot added eight connectors in preview. Story 3: Copilot Evals is in public preview for Individual, Business and Enterprise, and it lets you create a repeatable test for a skill, custom agent or MCP tool, approve each run, and compare results after a change. Ask which of our agents poll pull requests, which Copilot plan and teams we use, and whether we already customize skills or MCP tools, then give a short checklist for each story and link the official posts instead of inventing settings or flags.