The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes untrusted content from external websites, such as meta tags, titles, and sitemap entries. Maliciously crafted instructions within these fields could potentially influence the AI agent's behavior during the audit.
Ingestion points: Website content fetched via scripts/pages.mjs and parsed in scripts/inspect-metadata.mjs.
Boundary markers: None present to delimit external site content from agent instructions.
Capability inventory: The skill possesses network access, file system read access, and the ability to control a headless browser via Playwright.
Sanitization: Employs basic HTML parsing and entity decoding in scripts/html.mjs but lacks specific sanitization against prompt injection.
[DYNAMIC_EXECUTION]: The skill dynamically imports the playwright or playwright-core library at runtime using computed paths derived from the current working directory and local cache locations in scripts/lib/browser.mjs. While common for optional dependencies, this allows loading code from locations outside the skill's own package.
[COMMAND_EXECUTION]: The skill's primary functionality is implemented through Node.js scripts (inspect-metadata.mjs, check-links.mjs) that the agent is instructed to run, involving network and file system operations.
[EXTERNAL_DOWNLOADS]: The skill performs automated network requests to fetch metadata and verify links on external domains as specified by the user or discovered during the site audit.

seo-readiness