The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to Indirect Prompt Injection because it crawls and processes content from external URLs provided in the $ARGUMENTS.
Ingestion points: The skill uses a Node.js script with Playwright to fetch page.content() from an external website, which is then analyzed by the agent to identify UI elements.
Boundary markers: There are no instructions provided to wrap the fetched HTML content in delimiters or to ignore potential instructions embedded within the target page.
Capability inventory: The agent has access to Bash(node *), Bash(pnpm *), Bash(npx *), Write, and Edit, allowing it to execute arbitrary code or modify the local environment if coerced by malicious content on a crawled page.
Sanitization: No validation or sanitization of the external HTML content is mentioned before it is processed by the AI.
[COMMAND_EXECUTION]: The skill instructs the agent to execute arbitrary Node.js code via node -e to perform web scraping and to run a local CLI tool (dist/cli.js) for validation and execution of the generated YAML specs.
[EXTERNAL_DOWNLOADS]: The skill performs network operations by navigating to external, user-specified URLs using the Playwright library to discover application structures.

demo-from-url