reverse-ida-mcp-driver
Pass
Audited by Gen Agent Trust Hub on Oct 17, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted binary data from an IDA Pro instance using tools like find_regex, decompile, and py_eval, exposing the agent to potential instructions embedded in strings or symbols. Ingestion points: External binary data ingested via MCP tools. Boundary markers: Instructions for structured reporting and separating facts from inferences. Capability inventory: Binary modification (rename, set_type) and Python execution (py_eval). Sanitization: No explicit sanitization of extracted strings.
- [DYNAMIC_EXECUTION]: The skill relies on the py_eval MCP tool to execute Python scripts (scripts/ida_rodata_scanner.py, scripts/ida_panic_paths.py, scripts/ida_pclntab_finder.py) within the IDA Pro environment. These scripts are provided as local files and are wrapped in a standard IDA-specific block before execution. This constitutes dynamic execution of local code templates.
Audit Metadata