The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill fetches and analyzes rendered HTML from external websites using the bdata tool, creating a surface where malicious instructions hidden in a site's content could attempt to subvert the agent's audit or instructions.
Ingestion points: HTML content from sampled URLs saved to temporary files like /tmp/seo/_home.html (defined in references/bdata-recipes.md).
Boundary markers: The skill does not implement delimiters or safety instructions to prevent the agent from obeying instructions embedded in the analyzed web data.
Capability inventory: The agent can execute the bdata CLI and various shell utilities (grep, sed, python3).
Sanitization: No sanitization of the fetched HTML is performed before the agent processes the content.
[REMOTE_CODE_EXECUTION]: The skill instructions in SKILL.md include a command to install the Bright Data CLI via a remote script: curl -fsSL https://cli.brightdata.com/install.sh | bash. As this resource is owned by the skill's author and is a standard deployment method for their tool, it is documented as a functional requirement.
[DYNAMIC_EXECUTION]: In references/bdata-recipes.md (Recipe R-07), the skill dynamically generates and executes a Python script using a bash heredoc to parse JSON-LD schema from the fetched HTML files. This is a local data processing utility.
[EXTERNAL_DOWNLOADS]: The skill performs multiple external requests to fetch robots.txt, sitemap.xml, and rendered HTML pages using the bdata scrape command, as well as SERP data via bdata search.
[COMMAND_EXECUTION]: The skill relies on shell commands for data processing, including grep, sed, awk, and python3 to extract SEO metrics from HTML files.

seo-audit