electron-playwright-validator

Electron Playwright Validator

Use the bundled e-cli as the interface. It keeps one Electron process alive and connects subsequent commands through CDP, avoiding repeated startup cost.

Resolve the bundled tool

The executable is scripts/e-cli relative to this skill directory. Resolve that path from the host's skill context. In a full plugin install the equivalent path is:

E_CLI="${CLAUDE_PLUGIN_ROOT}/skills/electron-playwright-validator/scripts/e-cli"

Verify "$E_CLI" --help, then run from the project root. The tool resolves Electron and Playwright from the project's dependencies; do not install global copies or silently change the project's toolchain.

Direct skill installs can resolve scripts/e-cli from the skill directory. Use node "$E_CLI" when a distribution loses executable mode.

Drive one observable transaction at a time

1. Launch the app.
2. Read an accessibility snapshot.
3. Resolve one target from current state.
4. Act.
5. Wait for a product postcondition, then snapshot again.
6. Capture pixels only when they can reveal a rendering or layout defect.
7. Close the session when validation finishes.

Treat launch, interaction, and product success as separate claims. A successful click does not prove the view changed correctly.

Prefer semantic state

Use role and accessible-name selectors before text, test IDs, or CSS. The snapshot is both the discovery surface and the primary behavioral assertion. Load references/accessibility-selectors.md only when selector construction is non-obvious.

Electron rendering is asynchronous. Wait for the resulting element or state instead of relying on an arbitrary sleep. Snapshot again after navigation, modal changes, or asynchronous content. Read references/electron-gotchas.md for lazy rendering, terminal surfaces, and timing failures.

For multi-window apps, run e-cli pages and select the intended renderer rather than assuming the first window is the product surface.

Bound visual context