playwright-local
Pass
Audited by Gen Agent Trust Hub on Oct 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDATA_EXFILTRATIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to scrape and process content from external websites, which creates a potential surface for indirect prompt injection where malicious instructions on a webpage could influence the agent's behavior.
- Ingestion points:
templates/basic-scrape.tsandtemplates/stealth-mode.tsingest data viapage.goto()andpage.textContent(). - Boundary markers: Absent; scraped content is directly processed or logged by the agent.
- Capability inventory: File system access (
fs.writeFilefor screenshots and sessions) and command execution via the agent's environment (npx). - Sanitization: Absent; content is treated as raw text.
- [COMMAND_EXECUTION]: The skill includes a bash script
scripts/install-browsers.shthat executes package manager commands (npx playwright install) to download and configure browser binaries. This is a legitimate and documented setup procedure for the Playwright framework. - [DATA_EXPOSURE]: The
authenticated-session.tstemplate implements session persistence by saving browser cookies to a local file (session.json). While this is a standard requirement for maintaining login states in automation, users should be aware that these files contain sensitive session tokens. - [EXTERNAL_DOWNLOADS]: The skill facilitates the download of official browser binaries from Microsoft's Playwright infrastructure. It also utilizes well-known community packages like
playwright-extraandpuppeteer-extra-plugin-stealthto enhance anti-detection capabilities.
Audit Metadata