The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion of data from external, untrusted websites, creating a potential surface for indirect prompt injection attacks where malicious instructions on a webpage could influence the AI agent's subsequent actions.
Ingestion points: The scripts scripts/playwright-simple.js and scripts/playwright-stealth.js navigate to and extract content from arbitrary URLs provided as command-line arguments.
Boundary markers: The extracted content is returned as raw text or JSON output without structural delimiters or warnings that would instruct the agent to disregard instructions embedded in the scraped data.
Capability inventory: The skill has the ability to perform network requests, write files (screenshots and HTML) to the local disk, and execute commands via the Bash tool.
Sanitization: No sanitization, escaping, or filtering is applied to the extracted innerText or outerHTML before it is returned to the agent.
[EXTERNAL_DOWNLOADS]: The skill's setup process requires downloading external software components and binaries.
The installation instructions in INSTALL.md and README.md involve fetching the playwright package from the npm registry and downloading the Chromium browser binary from official Playwright distributions.
[COMMAND_EXECUTION]: The skill implements its functionality through the execution of Node.js scripts via a shell environment.
The agent is prompted to invoke the scripts using the Bash tool. While the scripts handle URL arguments via standard process argument arrays, the use of a general-purpose shell tool for automation expands the available attack surface for the agent.

playwright-scraper-skill