Skip to content

Suggested test: copilot/playwright-cli-mode #13518

Description

@github-actions

Motivation

Link to the gh-aw PRs: github/gh-aw#57521 (Remove built-in Playwright MCP support, default to CLI mode; mode: mcp now rejected) and github/gh-aw#57594 (Add AWF-aware Playwright CLI policy prompt enforcing loopback-only servers and Squid-routed egress). No existing test exercises tools: playwright in CLI mode at all.

Proposed test

  • Workflow file: test-copilot-playwright-cli.md
  • Trigger: workflow_dispatch
  • Engine: copilot
  • Safe output: create-issue (report a short result after running a local Playwright CLI check)
  • Variant: standard

Minimal test prompt sketch

Configure tools: { playwright: { mode: cli } } with the sandbox/firewall enabled. Ask the agent to start a local static server bound to 127.0.0.1, use the Playwright CLI to take a screenshot or snapshot of it, then create an issue summarizing success and confirming no external navigation occurred.

New fixtures or secrets needed

None — uses existing repo content for a static page; no new fixtures.

Notes

Verify mode: mcp truly fails to compile (could be asserted with gh aw compile in a lint-only check rather than an E2E dispatch test, since compile-time rejection isn't something workflow_dispatch proves). The CLI-mode happy path is the part suited to E2E; the mode:mcp rejection is more of a compiler unit-test concern.

Generated by 🔍 Suggest New E2E Tests · copilot · auto · 29.7 AIC · ⌖ 4.64 AIC · ⊞ 8.8K ·

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions