The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses shell commands to interact with the host operating system to open a browser.
Evidence: In pillar/pillar.ts, the openBrowser function uses child_process.exec to run platform-specific commands (open, start, xdg-open) with a URL constructed from dynamic data.
The URL includes an opId parameter fetched from the Pillar backend API. If the API returns a malicious payload containing shell metacharacters, it could lead to command injection, as the input is interpolated into a shell command string.
[CREDENTIALS_UNSAFE]: The skill implements a key management system that relies on environment variables for security.
Evidence: In pillar/pillar-direct.ts, the getDerivedPassword function generates a hash based on the PILLAR_API_KEY environment variable.
This derived hash is used as a password to unlock local secp256k1 signing keys stored in ~/.aibtc/signing-keys/. This design means that any entity with access to the environment variables can gain full access to the wallet's signing capabilities.

pillar