The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The documentation instructs users to provide a cryptographic private key via the FITTINGS_PRIVATE_KEY environment variable. Storing raw private keys in environment variables or shell history is a standard security risk, although the script uses this key only for local signing via the viem library.
[EXTERNAL_DOWNLOADS]: The skill documentation suggests fetching the necessary helper script call.mjs directly from the vendor's domain at https://fittings.sh/skill/call.mjs using curl if it is missing.
[COMMAND_EXECUTION]: The skill execution relies on node call.mjs, which performs network operations, accesses sensitive environment variables, and reads its own file content from the disk to perform a self-integrity check against a remote SHA-256 hash.
[REMOTE_CODE_EXECUTION]: The call.mjs script performs dynamic loading of Node.js modules (such as @x402/fetch, @x402/evm, and viem) at runtime using dynamic import() statements.
[PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface as it ingests untrusted input through the text parameter. This data is transmitted to the remote API without sanitization. Evidence:
Ingestion points: text parameter in SKILL.md and call.mjs CLI arguments.
Boundary markers: None present; text is passed directly into the JSON request body.
Capability inventory: Network POST requests via fetch to fittings.sh.
Sanitization: None; input is stringified directly into the request payload.

crypto-hash-digest