Web Qa Bot

AI-powered automated QA for web apps. Smoke tests, accessibility, visual regression. Works with Cursor, Claude, ChatGPT, Copilot. Vibe-coding ready.

This package appears coherent for automated web QA, but check a few practical points before installing or running it: - agent-browser and Playwright: The tool relies on the agent-browser CLI (peer dependency) and the lockfile references playwright-core; those packages may run install scripts and download browser binaries. That is normal for browser automation but expect large downloads and install-time activity. - Build/distribution mismatch: The repository content contains TypeScript source (src/) while the CLI bin refers to dist/cli.js. If you install a published package, verify the package includes a built dist/ directory; otherwise the CLI may fail to run. - File system and browser access: The tool writes screenshots and report files to disk and launches/controls browser instances. Do not point it at sensitive internal systems or provide secrets in test files unless you trust the package source. - PDF/reporting dependencies: SKILL.md mentions ai-pdf-builder and LaTeX for PDF export; those are not listed as direct dependencies — you may need to install extra tooling to generate PDFs. - Verify origin: The skill metadata lists a repo and npm name. If you plan to use this in production, verify the package on npm/GitHub (authors, recent releases, checksums) to avoid typosquat or forged packages. If you want, I can: (a) scan the omitted source files for any network endpoints or suspicious code paths, (b) check for any hard-coded URLs/credentials inside all files, or (c) produce a short checklist to safely run the first smoke test in an isolated environment.

Type: OpenClaw Skill Name: web-qa-bot Version: 0.1.3 The skill contains critical shell injection vulnerabilities. In `src/browser.ts`, user-controlled inputs such as URLs (in `goto`), selectors (in `click`, `hover`, `select`), screenshot names, and keyboard keys (in `press`) are directly concatenated into shell commands executed via `execSync` without sufficient sanitization. This allows an attacker to execute arbitrary commands on the host system by crafting malicious test suite files or CLI arguments. A similar vulnerability exists in `src/reporter.ts`, where `execSync` is used to invoke `ai-pdf-builder` with user-controlled output paths and company names, also without proper escaping.