omh-native-debugging
Pass
Audited by Gen Agent Trust Hub on Nov 9, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill defines a workflow for analyzing native binary faults without performing any direct execution or sensitive data access itself.
- [COMMAND_EXECUTION]: While the skill provides command-line templates for debuggers like GDB and LLDB, it contains explicit safety rules (e.g., 'Do not execute binaries, debuggers, or any command from OMH core') to ensure the agent only generates plans for a human or separate executor.
- [PRIVILEGE_ESCALATION]: Mentions of diagnostic tools that interface with system protections (e.g., ptrace, dtruss) are accompanied by warnings to only operate on binaries owned by the user, preventing unauthorized access attempts.
- [INDIRECT_PROMPT_INJECTION]: The skill accepts external diagnostic data (crash logs, sanitizer output) as input. It mitigates injection risks by separating the analysis phase from execution and requiring human/executor confirmation of all observed states.
Audit Metadata