The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it extracts and processes text content, metadata, and accessibility trees from untrusted external URLs.
Ingestion points: Untrusted data enters the agent context through the browser_navigate and browser_snapshot tools as described in Step 1 of the workflow in SKILL.md.
Boundary markers: The skill instructions do not specify the use of delimiters or 'ignore' instructions to prevent the agent from obeying commands embedded within the reference website's content.
Capability inventory: The skill has the capability to write new files (pages/*.vue, assets/content/*.json) and execute browser-based JavaScript.
Sanitization: There is no evidence of sanitization or validation of the extracted website content before it is used to drive the code generation process.
[COMMAND_EXECUTION]: The skill utilizes the browser_execute_javascript tool via the Playwright MCP server. While the script strings provided in the workflow (e.g., document.body.scrollHeight) are static and safe, the use of this capability on arbitrary user-provided pages is a design pattern that requires awareness.

reference-to-landing-page