The Agent Skills Directory

[DYNAMIC_EXECUTION]: The library scripts/lib/browser.mjs performs dynamic loading of dependencies (playwright or playwright-core) using createRequire and pathToFileURL. It also implements a discovery mechanism that searches for and attempts to execute Chromium binaries found in user-controlled environment variables (READYVIBE_CHROME) and common cache directories such as ~/.cache/ms-playwright.
[EXTERNAL_DOWNLOADS]: The script scripts/check-links.mjs makes outbound network requests using fetch to verify both internal and external links found on the target website. While consistent with the tool's purpose, this allows for interaction with arbitrary remote hosts if instructed or if found during a crawl.
[COMMAND_EXECUTION]: The skill relies on the execution of Node.js scripts (check-links.mjs) and manual curl commands to audit site behavior. The scripts further spawn headless browser processes to render pages.
[INDIRECT_PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by ingesting and parsing untrusted HTML content from target websites.
Ingestion points: scripts/check-links.mjs and scripts/lib/pages.mjs fetch and parse HTML from target URLs via fetchWithRetry and parseHtml.
Boundary markers: The findings generated from this data are reported to the agent without explicit boundary markers or instructions to disregard embedded commands.
Capability inventory: The skill has the capability to perform further network requests (fetch) and read local files (readFileSync).
Sanitization: While the parser extracts visible text via textOf, it does not sanitize that text for potential natural language instructions that could influence the agent's summary or next steps.

error-pages