The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external websites via the WebFetch tool, specifically processing PageSnapshot, rendered_dom, and raw_html. There is a risk that malicious instructions embedded in a target website's HTML could influence the agent's behavior.
Ingestion points: Processes rendered_dom and raw_html from external URLs in the Audit section.
Boundary markers: No explicit boundary markers or "ignore embedded instructions" warnings are defined for the ingested web content.
Capability inventory: The skill uses WebFetch, Bash, Read, Grep, and Glob tools, and executes shell commands for verification.
Sanitization: No sanitization or filtering of the external content is specified before processing.
[COMMAND_EXECUTION]: The skill utilizes the Bash tool to execute a local script for verification purposes.
Evidence: The Verification section defines a command: node ${CLAUDE_SKILL_DIR}/../../scripts/validate-jsonld.mjs --url <u>. This involves executing a Node.js script located at a relative path from the skill directory, passing a URL parameter that may be derived from untrusted input.

seo-eeat