Playwright Interactive

Persistent browser and Electron interaction through `js_repl` for fast iterative UI debugging.

Before installing or using this skill: (1) be aware the SKILL.md requires node/npm, Playwright (and optionally Electron) even though the registry metadata omitted those requirements — you'll run npm install which writes to disk. (2) The instructions explicitly tell you to run the agent with --sandbox danger-full-access and to modify ~/.codex/config.toml to enable js_repl; both grant the agent broad local access and should only be done in a trusted, isolated environment (e.g., disposable VM or container). (3) If you decide to proceed, review the project workspace you will run this from (avoid running in a directory that contains secrets), verify the Playwright package sources, and back up your ~/.codex/config.toml before changing it. (4) If you cannot run with an elevated sandbox safely, prefer not to enable this skill or replicate only the non-privileged parts of the workflow. Additional information that would change the assessment: an updated registry manifest that declares required binaries (node/npm) and a clear, minimal explanation why sandbox must be disabled or a safer alternative that works within the default sandbox.

Type: OpenClaw Skill Name: playwright-interactive Version: 1.0.0 The skill requires the user to explicitly disable sandboxing using the `--sandbox danger-full-access` flag and enable a JavaScript REPL, which grants the agent unconstrained access to the host system. While these permissions are plausibly needed for the stated purpose of persistent Playwright and Electron UI debugging, the requirement for full system access is a significant security risk. The SKILL.md file contains extensive JavaScript logic for session management and screenshot normalization that executes within this privileged environment.