ghidra-agent-cli

Warn

Audited by Gen Agent Trust Hub on Sep 22, 2026

Risk Level: MEDIUMDYNAMIC_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [DYNAMIC_EXECUTION]: The skill implements a function reconstruction and verification pipeline that automatically generates C test harnesses from decompiled code and compiles them using the system gcc compiler (src/rebuild.rs). It also dynamically generates and executes Frida JavaScript instrumentation scripts at runtime to capture and verify function I/O behavior (src/frida/scripts.rs, src/verify.rs).
  • [COMMAND_EXECUTION]: The utility frequently executes external system commands including gcc, frida, nm, ar, and Ghidra's analyzeHeadless binary to perform its core reverse engineering tasks.
  • [INDIRECT_PROMPT_INJECTION]: The tool is designed to ingest and process untrusted binary files and metadata artifacts. Although it includes logic to sanitize directory components and validate YAML paths, the automated processing of output from potentially malicious binaries creates a surface where binary data could influence agent-generated commands or analysis logic.
  • [EXTERNAL_DOWNLOADS]: The skill includes a standalone installation script (scripts/install-current-release.sh) that fetches binary archives directly from the official GitHub repository for the project. These downloads are from a well-known source and are used for the skill's own release management.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Sep 22, 2026, 03:47 AM