The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill requires the agent to execute a utility script call.mjs. If the script is not present, instructions direct the agent to download it from the vendor's domain via curl -sO https://fittings.sh/skill/call.mjs and execute it using Node.js.
[COMMAND_EXECUTION]: The documentation provides multiple bash commands for the agent to perform, including curl for fetching data and quotes, jq for parsing responses, and node for running the payment handling client.
[CREDENTIALS_UNSAFE]: The execution environment requires a FITTINGS_PRIVATE_KEY environment variable. The skill documentation explicitly notes that this key is stored in plaintext where the agent runs and recommends using a throwaway wallet holding minimal funds. This creates a risk of credential exposure if the agent's environment is compromised.
[INDIRECT_PROMPT_INJECTION]: The skill processes arbitrarily nested JSON data which is then flattened into dotted-path notation. This data ingestion represents a potential attack surface for indirect prompt injection if the output is interpreted by the agent without proper sanitization.
Ingestion points: Data is accepted through command-line arguments in call.mjs and parsed via parseArgs using the service's schema.
Boundary markers: None present; the script transforms the input based on the requested service schema.
Capability inventory: The script possesses network access (fetch) and the ability to load external libraries for cryptographic signing.
Sanitization: The script performs type validation (number, boolean, array, object) against the vendor's schema but does not perform content-level sanitization or escaping of the values processed.
[DYNAMIC_EXECUTION]: The call.mjs script utilizes dynamic import() to load dependencies such as @x402/fetch, @x402/evm, and viem at runtime.

fmt-json-flatten