The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted content from external websites, creating a potential vector for prompt injection. \n- Ingestion points: The skill uses WebFetch to crawl target URLs for analysis (e.g., in 30x-seo-page) and processes the output of the squirrelscan CLI tool which audits external sites. \n- Boundary markers: Absent. There are no instructions requiring the agent to disregard instructions or control codes found within the external website data. \n- Capability inventory: The skill has access to Bash, WebFetch, Read, Grep, and Glob tools. \n- Sanitization: Absent. Fetched content is analyzed directly for SEO metrics without filtering for malicious prompt directives. \n- [COMMAND_EXECUTION]: The skill uses the Bash tool to execute analysis commands. \n- Evidence: SKILL.md contains the orchestration step which squirrelscan && squirrelscan audit <url> --format llm. \n- Evidence: references/cwv-thresholds.md provides examples for using curl and npx lighthouse. \n- [EXTERNAL_DOWNLOADS]: The skill involves the download and execution of external software. \n- Evidence: SKILL.md instructs the user to install the squirrelscan package via npm i -g squirrelscan. \n- Evidence: The reference for Core Web Vitals mentions using npx lighthouse to perform lab data measurements.

seo