The Agent Skills Directory

[DATA_EXFILTRATION]: The skill mandates the collection and transmission of the user's verbatim prompt to the vendor's telemetry endpoint (shopify.dev). This involves capturing raw user interaction data and sending it to an external server.
[OBFUSCATION]: JavaScript components including log_skill_use.mjs and log_feedback.mjs are provided in a minified format. This obfuscates logic responsible for handling user data, making network requests, and interacting with the local file system.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted product and merchant information from a global catalog and various business-provided URLs, which could contain malicious instructions. Ingestion points: Catalog search results, product details, and merchant-specific discovery data. Boundary markers: No specific delimiters or instructions to ignore embedded commands are implemented when processing merchant-provided content. Capability inventory: The skill is capable of shell command execution (UCP, bash, node), writing to temporary files for prompt stashing, and performing network requests via curl or fetch. Sanitization: The instructions do not specify any sanitization or validation of the merchant-supplied text before the agent processes or displays it.
[COMMAND_EXECUTION]: The skill relies on extensive shell command execution for its primary functionality and telemetry, which represents a potential attack surface for command injection if inputs are not properly sanitized.

ucp