firmware-pentest
Warn
Audited by Gen Agent Trust Hub on Sep 4, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONPRIVILEGE_ESCALATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The installation instructions for the skill require cloning multiple third-party repositories and immediately executing their setup scripts with administrative privileges (e.g., executing sudo ./installer.sh within the emba repository).
- [PRIVILEGE_ESCALATION]: The skill directs the agent to use 'sudo' for a wide variety of tasks, including tool installation, system configuration for databases, mounting filesystems, and executing chroot environments.
- [EXTERNAL_DOWNLOADS]: The skill automates the fetching of numerous external tools and dependencies from GitHub, Cargo, and Pip during its 'On-Demand Bootstrap' phase, increasing the attack surface through unverified third-party code.
- [COMMAND_EXECUTION]: The primary function of the skill is to execute shell commands for firmware extraction, system emulation via QEMU, and running fuzzing tools like AFL++.
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted binary firmware images from the user. While a confirmation gate is included, there is no explicit sanitization for data passed into high-privilege subprocesses. Evidence chain: 1. Ingestion: firmware binaries (.bin, .img, .ubi). 2. Boundary: Confirmation gate present. 3. Capabilities: root access, network operations, file system manipulation. 4. Sanitization: Not explicitly defined for tool arguments.
Audit Metadata