The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted text and style data from external URLs, which creates a vector for hidden instructions in a website to influence the agent.
Ingestion points: Data is extracted from the DOM of external sites via page.evaluate calls within the scripts/extract.mjs script.
Boundary markers: No specific delimiters or "ignore" instructions are provided to the agent to distinguish between the extracted data and the agent's core instructions.
Capability inventory: The agent can execute a local script, write files (JSON and PNG) to the disk, and initiate network connections through a Playwright-controlled browser.
Sanitization: The skill lacks sanitization of the extracted web content before it is processed by the agent.
[COMMAND_EXECUTION]: The skill relies on the execution of a local Node.js script to perform its task.
Evidence: The workflow involves running node scripts/extract.mjs. The script includes a security check to prevent path traversal by validating that the output directory stays within the current working directory using path.resolve and path.relative logic.

extract-design-system