The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill is designed to extract and decrypt sensitive cookies from various browsers (Chrome, Edge, Brave, etc.). It explicitly attempts to access the macOS Keychain to retrieve the decryption keys required to read these credentials.\n- [COMMAND_EXECUTION]: The skill identifies and executes local binaries located in the current directory or the user's home directory. It runs subcommands like cookie-import-browser to perform its core functions.\n- [EXTERNAL_DOWNLOADS]: During setup, the skill downloads and executes an installation script from https://bun.sh/install. While this is a well-known source, it introduces an external dependency into the execution environment.\n- [REMOTE_CODE_EXECUTION]: The skill utilizes a curl | bash pattern to execute the Bun installer script directly from a remote URL during its setup phase.\n- [PROMPT_INJECTION]: The skill contains a vulnerability surface for indirect prompt injection via the --domain parameter.\n
Ingestion points: User input provided as a domain argument for the cookie import command.\n
Boundary markers: The skill does not use specific delimiters to isolate user-provided domain strings from the command execution context.\n
Capability inventory: The skill possesses the Bash tool permission, allowing it to execute arbitrary shell commands if the input is not sanitized.\n
Sanitization: There is no evidence of sanitization or escaping of the domain string before it is interpolated into the shell command.

setup-browser-cookies