devops(triage): add the issue triage and revision bots

.claude/skills/playwright-triage/SKILL.md

@@ -0,0 +1,85 @@
+---
+name: playwright-triage
+description: Triage and fix one microsoft/playwright-dotnet issue, or do a maintainer's task, unattended inside the Issue triage GitHub Actions workflow (.github/workflows/triage.yml), which a needs-triage, needs-triage-mac or needs-triage-win label starts, or a maintainer starts by hand with an issue, a task or both. Reproduces the issue on the runner, then leaves a verified fix proposal for the workflow to open as a draft PR, an investigation note for it to post on the issue, or an upstream bug report for a maintainer to file in microsoft/playwright. Only for that workflow.
+---
+
+# Playwright .NET Issue Triage
+
+You run unattended in the [Issue triage](../../../.github/workflows/triage.yml) workflow of
+`microsoft/playwright-dotnet`. First read the shared guide, `$AGENT_GUIDES_DIR/triage.md`, a copy of
+[microsoft/playwright/.github/actions/run-agent/guides/triage.md](https://github.com/microsoft/playwright/blob/main/.github/actions/run-agent/guides/triage.md),
+and then [CLAUDE.md](../../../CLAUDE.md), both in full. The guide says how to triage an issue in any
+Playwright repo. CLAUDE.md says how to build, test and format this one, and where a regression test
+goes. This file covers what neither has: the checkouts, how to run a published version, and the
+outcome that this repo adds, an upstream bug report.
+
+## The checkouts
+
+`$REPO_DIR` is the checkout of `microsoft/playwright-dotnet` that the guide describes. The driver is
+downloaded (`./build.sh --download-driver`), the solution is built (`dotnet build ./src`), and
+Chromium, Firefox and WebKit are installed
+(`pwsh src/Playwright/bin/Debug/netstandard2.0/playwright.ps1 install --with-deps`). It is the
+working directory, and the checkout that the guides refer to. The clone is shallow: run
+`git fetch --unshallow --tags origin` in it before you bisect or use `git tag --contains`. .NET SDK
+8 and 10 are installed, as in CI. Tests run with `-f net8.0`.
+
+`$PLAYWRIGHT_DIR` is a shallow clone of `microsoft/playwright` at `main`. The .NET client is a port
+of the JavaScript client in `packages/playwright-core/src/client/`. The API docs are in
+`docs/src/api/`, the API generator is `utils/doclint/generateDotnetApi.js`, and the upstream tests
+are in `tests/page/` and `tests/library/`. The clone is not at the commit of the driver:
+`<DriverVersion>` in `src/Common/Version.props` names that version, `main` is close to it, and a
+release has the tag `v<DriverVersion>`. The clone is not built. Before you run an upstream test in
+it, run `npm ci`, `npm run build` and `npx playwright install` there. Then `npm run ctest <spec>`,
+`npm run ftest <spec>` and `npm run wtest <spec>` run one spec file on Chromium, Firefox and WebKit.
+
+## Published versions
+
+To run a published version, for example to bisect, create
+`$SCRATCH_DIR/playwright-dotnet-<n>/<version>/` and make a console project there:
+
+```bash
+dotnet new console -n repro -f net8.0 -o . && dotnet add package Microsoft.Playwright --version <version>
+dotnet build && pwsh bin/Debug/net8.0/playwright.ps1 install chromium
+dotnet run
+```
+
+Put the repro in `Program.cs` as top-level statements, as `src/Playwright.Examples/Program.cs`
+does. `playwright.ps1 install` fetches the browsers of that version. Match the change to commits in
+`$REPO_DIR` with `git log v<good>..v<bad> -- src/Playwright/`, and to the JavaScript client in
+`$PLAYWRIGHT_DIR` with the same tags.
+
+## Upstream bugs
+
+The .NET client talks to the Node driver that `microsoft/playwright` publishes, and many bugs that
+are reported here are in the driver. That fix lands in `microsoft/playwright`. The family is one
+team, so this is routing, not a request to the reporter. Find the owner with the same calls in both
+clients: write the repro as a .NET test in `$REPO_DIR` first, then write the same calls as an
+upstream test in `$PLAYWRIGHT_DIR` and run it there. When the Node test fails the same way, the root
+cause is upstream. When only the .NET test fails, the bug is in the port: compare the method in
+`src/Playwright/Core/` with the same method in the JavaScript client, and fix it here.
+
+The interfaces in `src/Playwright/API/Generated/` and the types in
+`src/Playwright/Transport/Protocol/Generated/` are generated by `utils/doclint/generateDotnetApi.js`
+and `utils/generate_dotnet_channels.js` from the upstream docs and protocol. Never edit them by
+hand. A missing method, option or type is an upstream change, in `docs/src/api/` (the `langs`
+annotations decide what .NET gets) or in the generator. .NET-only overloads are hand-written in
+`src/Playwright/API/Supplements/`, and the implementations are in `src/Playwright/Core/`.
+
+For an upstream root cause, create `$OUTPUT_DIR/upstream/` with:
+
+- `report.md`: the report for a maintainer to file in `microsoft/playwright`, for an engineer who
+  did not read this issue: what fails, expected versus actual, the driver version and the browsers,
+  and the condensed Node test from `$PLAYWRIGHT_DIR` as the repro. Say that .NET shows the same.
+- `report.json`: `{"title": "..."}`, at most 256 characters.
+
+The comment job appends the report to your comment, for a maintainer to file. The bot never files
+it. Still write the comment as the note, without the report in it. Write `report.json` last, when
+`report.md` and the comment are done. Like `proposal.json`, it marks the report as ready.
+
+## PRs from third parties
+
+Do not run a PR that changes a `.csproj`, `.props` or `.sln` file, `nuget.config`, `build.sh`,
+`src/tools/Playwright.Tooling/` (the driver download), `src/Playwright/build/` (MSBuild targets that
+run on every build), `.github/`, `.azure-pipelines/`, `utils/`, `package.json`, `package-lock.json`,
+or the generated files under `src/Playwright/API/Generated/` and
+`src/Playwright/Transport/Protocol/Generated/`.

.github/actions/setup-triage-runner/action.yml

@@ -0,0 +1,36 @@
+name: 'Set up the triage runner'
+description: >
+  Prepares the environment that the triage and revision agents work in: the checkout built as CI
+  builds it, with the browsers and the Copilot CLI installed, and a checkout of microsoft/playwright
+  to read the client that this repo ports. Exports PLAYWRIGHT_DIR, the path of that checkout, and
+  CI_ENV_EXEC, the wrapper from microsoft/playwright that runs a command under xvfb on Linux and at
+  normal I/O priority on Windows.
+runs:
+  using: composite
+  steps:
+  - uses: microsoft/playwright/.github/actions/ci-env@main
+  - uses: actions/setup-dotnet@a98b56852c35b8e3190ac28c8c2271da59106c68 # v6.0.0
+    with:
+      dotnet-version: |
+        8.0.x
+        10.0.x
+  - uses: actions/setup-node@949feb2413d6458794dcd2491c4babbbce0c15c1 # v7.1.0
+    with:
+      node-version: '24'
+  - name: Download the driver and build
+    shell: bash
+    run: |
+      ./build.sh --download-driver
+      dotnet build ./src
+  - name: Install the browsers
+    shell: bash
+    run: pwsh src/Playwright/bin/Debug/netstandard2.0/playwright.ps1 install --with-deps
+  - name: Check out microsoft/playwright
+    shell: bash
+    run: |
+      dir="${RUNNER_TEMP//\\//}/playwright"
+      git clone --quiet --depth 1 https://github.com/microsoft/playwright "$dir"
+      echo "PLAYWRIGHT_DIR=$dir" >> "$GITHUB_ENV"
+  - name: Install the Copilot CLI
+    shell: bash
+    run: npm install -g @github/copilot

.github/workflows/revision.yml

@@ -0,0 +1,132 @@
+# Copilot addresses the maintainers' review comments on a draft PR that the triage workflow opened,
+# without write credentials. A fresh job pushes the new commit to the PR's branch using the app, so
+# CI runs on it. The steps that are the same for every Playwright repo are actions in
+# microsoft/playwright, and the shared guide they carry says how to revise; the prompt below has
+# what is specific to this repo.
+name: PR revision
+
+on:
+  # Apply needs-revision, needs-revision-mac or needs-revision-win; re-apply it to trigger another pass.
+  # pull_request_target always runs this file from main; the jobs only accept the triage app's own PRs.
+  pull_request_target:
+    types: [labeled]
+
+permissions: {}
+
+jobs:
+  prepare:
+    name: Collect the maintainers' comments
+    if: >-
+      github.repository == 'microsoft/playwright-dotnet' &&
+      contains(fromJSON('["needs-revision","needs-revision-mac","needs-revision-win"]'), github.event.label.name) &&
+      github.event.pull_request.state == 'open' &&
+      github.event.pull_request.user.login == 'microsoft-playwright-automation[bot]' &&
+      github.event.pull_request.head.repo.full_name == github.repository
+    runs-on: ubuntu-latest
+    permissions:
+      pull-requests: write
+    env:
+      PLATFORM: ${{ fromJSON('{"needs-revision":"linux","needs-revision-mac":"mac","needs-revision-win":"windows"}')[github.event.label.name] }}
+    outputs:
+      head: ${{ steps.collect.outputs.head }}
+      base: ${{ steps.collect.outputs.base }}
+      branch: ${{ steps.collect.outputs.branch }}
+      platform: ${{ env.PLATFORM }}
+    steps:
+    - name: Collect the PR and the maintainers' comments
+      id: collect
+      uses: microsoft/playwright/.github/actions/collect-review@main
+      with:
+        pull-request: ${{ github.event.pull_request.number }}
+    - name: Post the run link
+      env:
+        GH_TOKEN: ${{ github.token }}
+        PR: ${{ github.event.pull_request.number }}
+        PLATFORM_NAME: ${{ fromJSON('{"linux":"Linux","mac":"macOS","windows":"Windows"}')[env.PLATFORM] }}
+        RUN_URL: ${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}
+      run: |
+        body=$(printf "Hi, I'm the Playwright bot. I started to address the review comments on %s: [agent run](%s)\n<!-- playwright-revision-started -->" "$PLATFORM_NAME" "$RUN_URL")
+        gh pr comment "$PR" --body "$body" || echo "::warning::Could not post the run link"
+
+  revise:
+    name: Revise PR ${{ github.event.pull_request.number }} (${{ needs.prepare.outputs.platform }})
+    needs: prepare
+    runs-on: ${{ fromJSON('{"linux":"playwright-x64-ubuntu24-64-core","mac":"macos-15-xlarge","windows":"playwright-x64-server2025-64-core"}')[needs.prepare.outputs.platform] }}
+    timeout-minutes: 240
+    concurrency:
+      group: revision-${{ github.event.pull_request.number }}
+      cancel-in-progress: false
+    defaults:
+      run:
+        shell: bash
+    env:
+      PR: ${{ github.event.pull_request.number }}
+      HEAD_SHA: ${{ needs.prepare.outputs.head }}
+      BASE_SHA: ${{ needs.prepare.outputs.base }}
+      TRIAGE_PLATFORM: ${{ needs.prepare.outputs.platform }}
+    permissions:
+      copilot-requests: write
+    outputs:
+      has_revision: ${{ steps.agent.outputs.has-result }}
+    steps:
+    - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
+      with:
+        ref: ${{ needs.prepare.outputs.head }}
+        persist-credentials: false
+    - name: Fetch the base and record the directories
+      run: |
+        git fetch --no-tags --depth=1 origin "$BASE_SHA"
+        echo "OUTPUT_DIR=${RUNNER_TEMP//\\//}/revision" >> "$GITHUB_ENV"
+        echo "REVIEW_DIR=${RUNNER_TEMP//\\//}/review" >> "$GITHUB_ENV"
+        echo "SCRATCH_DIR=${RUNNER_TEMP//\\//}/scratch" >> "$GITHUB_ENV"
+        echo "REPO_DIR=${GITHUB_WORKSPACE//\\//}" >> "$GITHUB_ENV"
+        mkdir -p "${RUNNER_TEMP//\\//}/revision" "${RUNNER_TEMP//\\//}/scratch"
+    - uses: actions/download-artifact@9000827ccba6bdab643e8b6fd33ac0654aef8333 # v8.0.2
+      with:
+        name: review
+        path: ${{ runner.temp }}/review
+    - name: Set up the runner
+      uses: ./.github/actions/setup-triage-runner
+    - name: Revise with Copilot CLI
+      id: agent
+      # No GitHub tools and no URL access: the agent reads the PR only from the review the prepare job filtered.
+      uses: microsoft/playwright/.github/actions/run-agent@main
+      with:
+        prompt: |
+          Follow the shared guide revision.md and this repo's CLAUDE.md. Read both in full before starting. $PLAYWRIGHT_DIR is a shallow clone of microsoft/playwright at main, for reading the JavaScript client that this repo ports.
+
+          Pull request: https://github.com/microsoft/playwright-dotnet/pull/${{ github.event.pull_request.number }}
+        output-dir: ${{ env.OUTPUT_DIR }}
+        result-file: commit/commit.json
+        artifact-name: revision
+        exec: ${{ env.CI_ENV_EXEC }}
+
+  push:
+    name: Push the revision
+    needs: [prepare, revise]
+    # commit.json is written last, so a finished revision is pushed even when the job failed later.
+    if: ${{ !cancelled() && needs.revise.outputs.has_revision == 'true' }}
+    runs-on: ubuntu-latest
+    steps:
+    - uses: actions/create-github-app-token@64ab729a6294000f5aa4a4bc0b73d1f32bc45510 # v3.3.0
+      id: app-token
+      with:
+        app-id: ${{ vars.PLAYWRIGHT_APP_ID }}
+        private-key: ${{ secrets.PLAYWRIGHT_PRIVATE_KEY }}
+    - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
+      with:
+        ref: ${{ needs.prepare.outputs.head }}
+        token: ${{ steps.app-token.outputs.token }}
+    - uses: actions/download-artifact@9000827ccba6bdab643e8b6fd33ac0654aef8333 # v8.0.2
+      with:
+        name: revision
+        path: ${{ runner.temp }}/revision
+    # Never execute code or scripts from the proposed changes in the app job.
+    - name: Validate and push
+      uses: microsoft/playwright/.github/actions/publish-patch@main
+      with:
+        token: ${{ steps.app-token.outputs.token }}
+        dir: ${{ runner.temp }}/revision/commit
+        patch: revision.patch
+        metadata: commit.json
+        branch: ${{ needs.prepare.outputs.branch }}