binary-protection-bypass

Warn

Audited by Gen Agent Trust Hub on Jul 21, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill provides instructions to execute commands with elevated privileges to modify system-wide security settings.
  • Evidence: The file PROTECTION_BYPASS_MATRIX.md contains the command echo 0 | sudo tee /proc/sys/kernel/randomize_va_space, which allows the agent to disable Address Space Layout Randomization (ASLR) globally at the kernel level.
  • [COMMAND_EXECUTION]: The skill relies on the execution of various system analysis and development tools for its primary functionality.
  • Evidence: Instructions frequently reference the use of checksec, readelf, and gcc for auditing binary protections and environment preparation.
  • [PROMPT_INJECTION]: The skill design creates an ingestion point for indirect prompt injection by processing external binary files.
  • Ingestion points: External binary files analyzed via checksec and readelf (referenced in SKILL.md).
  • Boundary markers: Absent. The instructions do not define delimiters or warnings for the agent when processing data from these binaries.
  • Capability inventory: The skill utilizes shell command execution, file system access, and privileged command execution (sudo).
  • Sanitization: Absent. There is no mention of sanitizing or validating binary metadata or strings before they are processed by the agent.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Jul 21, 2026, 01:23 PM
Security Audit — agent-trust-hub — binary-protection-bypass