The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a local script (check-answerblocks.mjs) relative to the skill directory to perform word counts and structural audits. This is a standard use of helper scripts bundled with the skill.
[EXTERNAL_DOWNLOADS]: The skill uses the WebFetch tool to retrieve rendered_dom or raw_html from user-specified URLs. This network activity is restricted to the intended purpose of auditing external SEO content.
[PROMPT_INJECTION]: The skill processes untrusted web content (Indirect Prompt Injection surface). Maliciously crafted web pages could attempt to influence the audit results or proposed rewrites, though the skill instructs the agent to surface items one at a time for user acceptance, providing a human-in-the-loop safeguard.
Ingestion points: rendered_dom and raw_html retrieved from external URLs (SKILL.md).
Boundary markers: None specified for the untrusted content.
Capability inventory: Bash tool used to execute a local Node.js audit script (SKILL.md).
Sanitization: None specified; relies on user review of proposed content rewrites.

seo-geo-answerblocks