The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data by fetching and analyzing the HTML and rendered DOM of external web pages. If an audited page contains malicious instructions (e.g., hidden in HTML comments or metadata), it could potentially influence the agent's behavior during the audit process.
Ingestion points: Processes rendered_dom and raw_html from external URLs via WebFetch as described in SKILL.md.
Boundary markers: The skill does not define explicit delimiters or instructions to the agent to treat the audited HTML content strictly as data and to ignore any instructions embedded within it.
Capability inventory: The skill utilizes Bash, WebFetch, Grep, Glob, and Read tools to perform its tasks.
Sanitization: The skill body does not specify any sanitization, filtering, or validation steps for the fetched HTML before it is presented to the agent for analysis.
[COMMAND_EXECUTION]: The skill performs verification by executing a Node.js script through the shell using the Bash tool. This command incorporates a user-supplied URL as an argument.
Evidence: node ${CLAUDE_SKILL_DIR}/../../scripts/psi-client.mjs --url <u> in SKILL.md.

seo-core-web-vitals