funnel-audit
Fail
Audited by Gen Agent Trust Hub on Aug 11, 2026
Risk Level: HIGHCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONDATA_EXFILTRATION
Full Analysis
- [COMMAND_EXECUTION]: The instruction to execute
/bin/rmbc-workspace active "<user's answer>"introduces a command injection vulnerability. If a user provides input containing shell metacharacters such as semicolons or command substitutions, it could lead to the execution of arbitrary commands within the agent's environment. - [COMMAND_EXECUTION]: The skill executes various local binaries and shell utilities (
rmbc-update-check,rmbc-analytics,rmbc-workspace,mkdir,touch,sed) to manage environment state and persistent configuration files in~/.rmbc-skills/. - [REMOTE_CODE_EXECUTION]: The skill implements a dynamic upgrade process that directs the agent to read and follow instructions from external markdown files (
skills/rmbc-upgrade/SKILL.md) at runtime. This allows for the modification of agent behavior based on the contents of these external files. - [DATA_EXFILTRATION]: Usage statistics are logged via a dedicated tool (
rmbc-analytics). Although the skill documentation states data remains local, this process involves executing a binary specifically designed to handle telemetry related to skill and product usage.
Recommendations
- AI detected serious security threats
Audit Metadata