The Agent Skills Directory

[CREDENTIALS_UNSAFE] (HIGH): The documentation specifies a --password global flag for providing device credentials via the CLI (references/user-interface.md). This practice is insecure as passwords can be exposed in process lists, shell history, and logs.
[REMOTE_CODE_EXECUTION] (HIGH): The skill provides instructions for the system execute command (references/limitations.md) and firmware install (references/quick-start.md). These allow for the execution of arbitrary code and the installation of new operating systems on remote network devices.
[COMMAND_EXECUTION] (MEDIUM): The tool is designed to execute various system-level operations including network scanning, system reboots, and service management on mining hardware.
[DATA_EXFILTRATION] (LOW): The tool can write data to local files via the --output and --logfile-path options, which could be misused if the agent is misdirected to write to sensitive locations.
[INDIRECT_PROMPT_INJECTION] (HIGH): Vulnerability surface identified: 1. Ingestion points: scan command (network) and --ip-file (local files); 2. Boundary markers: Absent; 3. Capability inventory: system execute and firmware install (references/limitations.md, references/quick-start.md); 4. Sanitization: Not documented. Malicious data from scanned devices could influence high-privilege agent actions.

braiins-toolbox