The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external websites to generate Next.js components and configuration.
Ingestion points: Hydrated dom.html, headings.json, and CSS/JS files mirrored to the local filesystem during the capture phase (documented in references/foundations-agent.md).
Boundary markers: The skill instructions do not specify the use of delimiters or specific warnings to the model to ignore instructions embedded within the scraped content.
Capability inventory: The skill executes shell scripts (scripts/), writes a full project structure to the filesystem (nextjs-project/), and performs network operations via browser automation.
Sanitization: There is no evidence of sanitization, escaping, or filtering of the ingested HTML/CSS content before it is interpolated into generated templates.
[COMMAND_EXECUTION]: The script scripts/capture-url.sh executes a shell command string passed via the --browser-command argument or BROWSER_CAPTURE_CMD environment variable using sh -c. This pattern allows for the execution of arbitrary shell commands if the command string is influenced by malicious input.
[DYNAMIC_EXECUTION]: The skill's primary objective is the runtime generation of a complete Next.js project based on untrusted external input. The resulting code, which includes components and configuration, is intended to be executed in the user's local environment through standard build tools like npm or npx.

convert-url-to-nextjs