moos-app-builder
Pass
Audited by Gen Agent Trust Hub on Sep 20, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes project files and environment variables to influence its code generation and command execution workflows.
- Ingestion points: The agent resolves the MOOS_IVP_ROOT from the environment and searches the user's workspace for existing build configurations (CMakeLists.txt).
- Boundary markers: The skill instructs the agent to validate the MOOS-IvP checkout structure and ask for user confirmation if paths are ambiguous.
- Capability inventory: The skill can write source code, modify build files, and execute shell commands including the framework's generator and the pAntler process manager.
- Sanitization: There is no evidence of robust sanitization for paths interpolated into shell commands beyond basic existence checks.
- [DYNAMIC_EXECUTION]: The skill follows a workflow that involves generating C++ source code using framework scripts, building the code via cmake, and then executing the compiled binaries (using --help, --example, and --interface flags) as smoke checks. It also executes an internal shell script (check_portable_paths.sh) provided with the skill.
- [COMMAND_EXECUTION]: The agent is tasked with running several shell-based tools including GenMOOSApp_AppCasting and pAntler. These executions are dependent on the resolved path for MOOS_IVP_ROOT; if an attacker controls this path or the environment, they could potentially induce the execution of malicious scripts.
Audit Metadata