The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a local script parse_html.py using the ${CLAUDE_PLUGIN_ROOT} environment variable to process URLs and extract structured data. This is a standard and safe use of local tooling within the agent's provided environment.
[DATA_EXPOSURE]: The skill processes user-supplied URLs to extract public SEO metadata. It does not access sensitive local files or credentials. Example templates use standard placeholders like [Company Name] and [Phone] which are safe.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted HTML data from external URLs via the parse_html.py script. While this presents an attack surface for instructions hidden in HTML comments or tags, the skill's primary function is data extraction (JSON-LD) rather than executing instructions found within that data. Sanitization is performed by the parsing script before the content reaches the LLM context.

seo-schema