prowl-cli
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses the
prowlCLI tool andjqutility to perform operations such as listing panes, reading agent states, and sending commands to sibling processes. These operations are restricted to the local environment and the application's own context. - [INDIRECT_PROMPT_INJECTION]: The skill defines a surface for processing data from external agent sessions, which is inherent to its purpose of session management. \n
- Ingestion points: External agent data enters the context via
prowl agents read,prowl read, andprowl list(SKILL.md). \n - Boundary markers: The skill documentation explicitly recommends the use of the
--jsonflag andjqfor structured parsing to ensure that agent output is treated as data rather than instructions. \n - Capability inventory: The skill has the ability to send text and keys to other panes via
prowl sendandprowl key(SKILL.md). \n - Sanitization: The skill advises using shell variable quoting (via
jq --argand single quotes for payloads) and identity verification using$PROWL_PANE_IDto minimize risks of acting on incorrect or unverified context.
Audit Metadata