iida-mcp-ida-integration

Warn

Audited by Gen Agent Trust Hub on May 16, 2026

Risk Level: MEDIUMEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill references an external Windows kernel driver (iida-mcp-ioctl.sys) and instructions for installing the capstone library via pip for disassembly features.
  • [COMMAND_EXECUTION]: Provides tools like patch_bytes that allow the agent to modify binary data within the IDA Pro database, and describes commands for installing and starting a system service (sc create, sc start) for the kernel driver component.
  • [DATA_EXFILTRATION]: The skill exposes powerful memory-reading tools (read_bytes, kernel_read_memory) and file metadata tools (get_file_info) which could be used by a malicious agent to exfiltrate binary contents or kernel memory. By default, the server listens on 0.0.0.0, exposing these capabilities to the network without authentication.
  • [PROMPT_INJECTION]: Vulnerable to indirect prompt injection (Category 8) as it processes untrusted binary data from IDA Pro.
  • Ingestion points: Raw bytes, strings, and decompiled code from binary files loaded in IDA Pro (SKILL.md).
  • Boundary markers: None identified in tool outputs like search_text or decompile to prevent the agent from following instructions embedded in binary strings.
  • Capability inventory: Memory reading (read_bytes), memory writing (patch_bytes), and kernel memory access (kernel_read_memory).
  • Sanitization: No specific sanitization or filtering of analyzed content is mentioned before it reaches the agent context.
Audit Metadata
Risk Level
MEDIUM
Analyzed
May 16, 2026, 11:21 PM
Security Audit — agent-trust-hub — iida-mcp-ida-integration