The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill instructs users to export a private blockchain key (FITTINGS_PRIVATE_KEY) into their environment. While it correctly warns users to use a throwaway wallet, storing private keys in plaintext environment variables is a high-risk practice that exposes credentials to any process or agent with environment access.
[COMMAND_EXECUTION]: The skill relies on the execution of call.mjs via node to perform its primary function. This script constructs shell commands and performs network requests based on user input.
[EXTERNAL_DOWNLOADS]: The documentation provides a curl command to fetch the call.mjs script directly from https://fittings.sh/skill/call.mjs if it is missing. While this targets the vendor's domain, downloading and executing scripts from remote sources at runtime is a notable security surface.
[REMOTE_CODE_EXECUTION]: The call.mjs script dynamically imports several external libraries (@x402/fetch, @x402/evm, viem) at runtime. It also performs a self-integrity check by fetching a SHA256 hash from a remote server (https://fittings.sh/skill/call.mjs.sha256) and comparing it against its own local hash, which gates execution.
[INDIRECT_PROMPT_INJECTION]:
Ingestion points: The skill processes US tariff codes provided as user input (code field).
Boundary markers: None identified in the provided scripts; input is coerced to types but not sanitized against prompt injection patterns.
Capability inventory: The skill has the capability to perform network requests (fetch), read the local file system (readFileSync), and sign blockchain transactions using a provided private key.
Sanitization: Minimal; the script performs type coercion (e.g., ensuring numbers are finite) but does not validate the content of string inputs against malicious instructions.

hts-lookup