reverse-engineer
Fail
Audited by Gen Agent Trust Hub on Nov 1, 2026
Risk Level: HIGHPRIVILEGE_ESCALATIONREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [PRIVILEGE_ESCALATION]: Instructions for installing system dependencies using sudo apt install require the agent to execute commands with root privileges.
- [PRIVILEGE_ESCALATION]: Recommends global installation of the rea-agents package (npm install --global), which typically requires elevated permissions and modifies system-wide directories.
- [REMOTE_CODE_EXECUTION]: Utilizes npx to download and execute packages such as @electron/asar and rea-agents dynamically at runtime.
- [EXTERNAL_DOWNLOADS]: Fetches various tools and packages from external sources including Homebrew, APT registries, and NPM.
- [EXTERNAL_DOWNLOADS]: References the official Ghidra website (https://ghidra-sre.org) for downloading analysis engines.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted binary data and source code. 1. Ingestion points: outputs from strings, otool, and rea commands. 2. Boundary markers: absent. 3. Capability inventory: shell execution (otool, binwalk, rea, npx), file system access (cp, realpath). 4. Sanitization: absent.
- [COMMAND_EXECUTION]: Employs a wide range of utilities including nm, objdump, binwalk, tshark, and lipo to investigate binary files and system configurations.
Recommendations
- AI detected serious security threats
Audit Metadata