The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill instructs the user to set a cryptocurrency private key (FITTINGS_PRIVATE_KEY) as an environment variable to authorize payments. While the documentation provides warnings and suggests using a throwaway wallet, storing private keys in environment variables creates a risk of exposure to other processes or system logs.
[EXTERNAL_DOWNLOADS]: The skill requires downloading the call.mjs script from the vendor's domain (fittings.sh) to function. It also fetches a SHA256 hash from the same domain to verify the script's integrity before execution.
[REMOTE_CODE_EXECUTION]: The call.mjs script uses dynamic imports to load external Node.js packages (@x402/fetch, @x402/evm, viem) at runtime. These packages are used to facilitate the blockchain-based payment handshake.
[COMMAND_EXECUTION]: The skill requires the execution of shell commands (node call.mjs) to interact with the URL expansion service.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-provided URLs and returns the final destination content. This presents a potential attack surface where the destination page could contain instructions intended to influence the agent's subsequent behavior.
Ingestion points: The url parameter passed to the expansion service in SKILL.md.
Boundary markers: None identified in the provided instructions to isolate external content from the agent's prompt.
Capability inventory: The skill allows network requests via fetch and execution of local scripts via node (in call.mjs).
Sanitization: No specific sanitization or filtering logic is mentioned for the content retrieved from the expanded URLs.

url-expand