The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the Bash tool to execute a local verification script (check-answerblocks.mjs) using Node.js. The command structure node ${CLAUDE_SKILL_DIR}/../../scripts/check-answerblocks.mjs --url <u> indicates execution of a component located in the environment's shared scripts directory.
[INDIRECT_PROMPT_INJECTION]: The skill's primary function involves ingesting and analyzing untrusted external data (rendered_dom, raw_html) from arbitrary URLs to generate content audits and fix suggestions. This provides a surface for indirect prompt injection if the external content contains instructions intended to manipulate the agent's output.
Ingestion points: Processes external HTML and DOM content fetched via the WebFetch tool.
Boundary markers: The skill specifies a manual_review verification method, ensuring a human review of the generated 'proposed' fixes, which serves as a safety boundary.
Capability inventory: The skill possesses the ability to execute shell commands (Bash), read/list files (Read, Glob, Grep), and fetch remote content (WebFetch).
Sanitization: The instructions do not describe explicit sanitization or filtering of the ingested HTML content before it is passed to the analysis logic or the verification script.
[DYNAMIC_EXECUTION]: The skill uses Bash to invoke the Node.js runtime for executing a script at a relative path. While the script is local, this pattern of dynamic execution is used to perform the core logic of the skill.

seo-geo-answerblocks