debugging-livekit-agents
Pass
Audited by Gen Agent Trust Hub on Oct 3, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to interact with the local environment by executing shell commands via the
lkCLI tool (e.g.,lk agent debugger start,lk agent debugger say). This is consistent with the skill's primary purpose of local agent development and testing. - [INDIRECT_PROMPT_INJECTION]: The skill creates a surface for indirect prompt injection by taking arbitrary user-supplied conversation turns and interpolating them into shell commands.
- Ingestion points: User-provided strings used as arguments for the
lk agent debugger saycommand inSKILL.md. - Boundary markers: Absent; the instructions use basic double-quoting in examples but do not provide robust delimiters or warnings to ignore instructions within the user turns.
- Capability inventory: The skill utilizes shell command execution capabilities to interact with the LiveKit debugger.
- Sanitization: Absent; there are no instructions provided to sanitize, escape, or validate the user-supplied text before it is passed to the system shell.
Audit Metadata