The Agent Skills Directory

[COMMAND_EXECUTION]: The skill launches a local browser instance (Chrome, Edge, or Chromium) via child_process.spawn in scripts/cdp.ts. This is used to render the target URL and handle JavaScript-heavy sites.
[DYNAMIC_EXECUTION]: The skill uses the Chrome DevTools Protocol (Runtime.evaluate) in scripts/cdp.ts to execute a large block of JavaScript within the target website's context. This script (cleanupAndExtractScriptBody in scripts/html-to-markdown.ts) is used to extract metadata and clean the DOM before conversion.
[INDIRECT_PROMPT_INJECTION]: The skill processes arbitrary content from the internet. Maliciously crafted websites could contain instructions designed to influence or override the agent's behavior when the resulting markdown is processed.
Ingestion points: The captureUrl function in scripts/main.ts fetches content from user-provided URLs.
Boundary markers: Absent. The resulting markdown is returned to the agent without explicit delimiters or instructions to ignore embedded commands.
Capability inventory: The skill can spawn subprocesses (browser), write to the local file system (saving markdown files), and perform network operations via the browser.
Sanitization: The skill uses Readability and Turndown to strip scripts and styles, and includes a sanitizeHtml function. However, it does not filter natural language instructions that could trigger injection.
[EXTERNAL_DOWNLOADS]: The SKILL.md file recommends using npx -y bun, which will download the Bun runtime if it is not already present in the user's environment.

baoyu-url-to-markdown