v0.1.10

1b5d/llm-apiv0.1.10Apr 30, 2026by yury-s

AI Summary

Refactors network inspection commands into numbered parts and introduces a spec-driven testing skill.

Key Highlights

  • Network inspection commands split: `network` replaced by `requests`, `request <num>`, and body/header extractors.
  • New spec-driven testing skill added to guide agents through plan/generate/heal workflows.
  • Read-only data commands now emit raw output by default.
  • Extension mode now ignores config file browser settings (userDataDir, executablePath).

Breaking Changes

  • Renamed `network` command to `requests` and added sub-commands
  • Read-only commands now emit raw output by default
  • Extension mode ignores browser settings from config files

New Features

  • Spec-driven testing skill
  • Refactored network inspection commands

Full Release Notes

## Highlights

- **Network inspection split into numbered commands** ([#377](https://github.com/microsoft/playwright-cli/issues/377)) — `network` is replaced with:
  - `requests` — numbered list of requests with stable indexes
  - `request <num>` — full details for one request
  - `request-headers <num>`, `request-body <num>`, `response-headers <num>`, `response-body <num>` — pipe-friendly part extractors

  Bodies are no longer inlined; pass `--filename` on any part-command to save the result to a file. ([microsoft/playwright#40447](https://github.com/microsoft/playwright/pull/40447), [microsoft/playwright#40454](https://github.com/microsoft/playwright/pull/40454))

- **Spec-driven testing skill** — a new `references/spec-driven-testing.md` reference guides agents through a plan / generate / heal workflow for driving Playwright tests from a written spec. ([microsoft/playwright#40460](https://github.com/microsoft/playwright/pull/40460))

## Behavior changes

- Read-only data-fetching commands (`cookie-list`, `cookie-get`, `localstorage-list`, `localstorage-get`, `sessionstorage-list`, `sessionstorage-get`, `route-list`, `request-headers`, `request-body`, `response-headers`, `response-body`) now emit raw output by default — no more `### Result` wrapper, no `--raw` needed. ([microsoft/playwright#40473](https://github.com/microsoft/playwright/pull/40473))
- Extension mode now ignores the `browser` section of any loaded config file (both `userDataDir` and `executablePath`); only the browser/channel from `--browser` or `PLAYWRIGHT_MCP_BROWSER` applies. ([microsoft/playwright#40475](https://github.com/microsoft/playwright/pull/40475))
- `requests` now appends a hint when static requests are filtered out, telling the user to pass `--static` to see them. ([microsoft/playwright#40454](https://github.com/microsoft/playwright/pull/40454))

## Fixes

- `fix(cli): resolve initPage/initScript paths and surface load errors` ([#290](https://github.com/microsoft/playwright-cli/issues/290)) — relative `initPage` / `initScript` paths from `--config` files now resolve against the config dir (matching Vite/Vitest/ESLint), and CLI-flag / env-var entries resolve against `cwd`. Load errors are no longer silently swallowed. ([microsoft/playwright#40451](https://github.com/microsoft/playwright/pull/40451))
- `fix(mcp): detect extension in non-default Chrome profiles` — the `--extension` preflight now scans every `Profile *` subdirectory under the user data dir, not just `Default`. ([microsoft/playwright#40471](https://github.com/microsoft/playwright/pull/40471))
- `fix(mcp): surface unhandled rejections instead of crashing the server` — async errors from user-installed callbacks (e.g. a `page.route` handler) are now reported on the next tool response instead of tearing down the MCP transport. ([microsoft/playwright#40452](https://github.com/microsoft/playwright/pull/40452))