playwright-scraper-skill
Pass
Audited by Gen Agent Trust Hub on Nov 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill serves as a primary ingestion point for untrusted content from the internet, which creates a significant surface for indirect prompt injection attacks.
- Ingestion points: The scripts
scripts/playwright-simple.jsandscripts/playwright-stealth.jsuse Playwright to navigate to and extract text content from any URL provided to the skill. - Boundary markers: Scraped content is returned as raw text within a JSON object without any protective delimiters or explicit instructions for the agent to ignore embedded commands.
- Capability inventory: The skill possesses network access (via Playwright) and file system write capabilities (
fs.writeFileSyncandpage.screenshot). This combination is sensitive as it allows the agent to process external data and then act upon the local environment. - Sanitization: No sanitization, filtering, or HTML-entity escaping is performed on the scraped content before it is passed to the AI agent's context.
- [COMMAND_EXECUTION]: The skill enables the execution of browser automation scripts that interact with the local file system and network.
- The
scripts/playwright-stealth.jsscript allows defining aSCREENSHOT_PATHvia environment variables. If an attacker can influence the agent to set this path, it could be used to write files to arbitrary locations or overwrite existing files. - The skill uses browser arguments such as
--no-sandboxand--disable-setuid-sandbox, which are common for running browsers in containerized environments but reduce the security isolation of the browser process.
Audit Metadata