The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill fetches an installation script for the Bright Data CLI from https://cli.brightdata.com/install.sh, which is an official resource provided by the author.
[REMOTE_CODE_EXECUTION]: The CLI tool is installed by piping a remote script directly to the bash shell (curl -fsSL ... | bash). This behavior is restricted to the vendor's own verified infrastructure.
[INDIRECT_PROMPT_INJECTION]: The skill ingests HTML content from arbitrary external websites to perform audits, which could expose the agent to malicious data designed to influence its reasoning or output.
Ingestion points: Data is retrieved from user-specified URLs and sitemaps using bdata scrape as described in references/bdata-recipes.md.
Boundary markers: The agent follows defined extraction recipes which provide structural isolation, though these do not fully eliminate injection risks.
Capability inventory: The agent has access to web scraping and shell command execution.
Sanitization: Processing is performed using standard utilities like grep, jq, and python3 to extract specific SEO elements.
[DYNAMIC_EXECUTION]: The skill executes a Python script provided via a heredoc in references/bdata-recipes.md (R-07) to extract JSON-LD schema markup from rendered HTML.
[COMMAND_EXECUTION]: The audit workflow utilizes standard command-line tools including grep, sed, awk, and jq for data analysis.

seo-audit