The Agent Skills Directory

Credential Handling Considerations: The skill describes the use of --private-key and --password flags, as well as environment variables like PURL_PASSWORD. While these are functional requirements for the purl tool's wallet management, passing sensitive credentials directly via command-line arguments or environment variables is a pattern to review, as they may be visible in process lists or shell history.
Financial Transaction Surface: The skill enables the agent to perform financial transactions (x402 payments) automatically. While the documentation highlights safety features like --max-amount and --confirm, there is an inherent risk if the agent processes untrusted URLs that trigger unintended payments.
Command Execution: The skill relies on executing the purl CLI tool to perform its primary functions. This involves subprocess calls with user-provided arguments such as URLs, headers, and data payloads.
Indirect Prompt Injection Surface:
Ingestion points: The skill accepts external input via the <URL> argument and data payloads (--data, --json).
Boundary markers: There are no explicit boundary markers or delimiters mentioned to separate untrusted URL content from the agent's internal instructions.
Capability inventory: The purl tool has the capability to perform network operations and execute financial transactions on EVM and Solana networks.
Sanitization: The skill provides safety parameters such as --max-amount to limit financial exposure and --confirm to require human-in-the-loop validation before a transaction is finalized.

pay-for-http-request