audio-hooks
Pass
Audited by Gen Agent Trust Hub on Nov 6, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill operates entirely through the
audio-hooksCLI binary, executing subcommands for configuration, testing, and diagnostics based on user requests. - [EXTERNAL_DOWNLOADS]: The skill instructs the agent to install plugins and update the CLI tool from the developer's marketplace and repository (ChanMeng666/echook).
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-provided values for configuration keys, durations, and webhook URLs, which are then passed to the command line, creating a vulnerability surface for malformed inputs.
- Ingestion points: User requests for notification preferences, snooze durations, and webhook URLs captured in the
Mapping a requestsection of SKILL.md. - Boundary markers: None identified; instructions do not specify how to escape or delimit user input passed to the CLI.
- Capability inventory: The skill can modify configuration files, enable/disable hooks, and perform network operations (webhooks) using the
audio-hooksCLI in SKILL.md. - Sanitization: Not present in the prompt; the system relies on the underlying binary to handle input validation.
- [PROMPT_INJECTION]: The instructions explicitly direct the agent to bypass user confirmation for all commands ("Do not ask the user to confirm... If they gave an instruction, run it"). This directive removes a standard safety oversight mechanism and could be exploited to perform unintended configuration changes or data restoration if the agent is manipulated.
Audit Metadata