The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill fetches content from external, untrusted URLs to conduct audits. This process involves ingesting untrusted data which is a standard vector for indirect prompt injection.
Ingestion points: Target website HTML and structured data are fetched by scripts/fetch-page.py, scripts/check-page.py, scripts/check-site.py, and scripts/check-schema.py.
Boundary markers: There are no explicit delimiters or instructions provided to the agent to distinguish between its operational instructions and the potentially hostile content of the audited pages.
Capability inventory: The skill is capable of performing network requests, writing files to the reports/ directory, and executing the open command via the agent.
Sanitization: The skill does not include automated mechanisms to sanitize extracted text before the agent embeds it into the HTML report template. This could lead to stored cross-site scripting (XSS) in the generated audit files.
[COMMAND_EXECUTION]: The skill instructions suggest that the agent executes the open command to display the generated HTML report to the user. The file path for this command is constructed based on the hostname of the target URL.
[EXTERNAL_DOWNLOADS]: The skill requires the requests Python library. requests is a widely-used and trusted library for HTTP communication.

seo-audit