axiom-lldb
Pass
Audited by Gen Agent Trust Hub on Oct 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill provides instructional content for runtime debugging using standard LLDB commands. It does not contain any executable code, network operations, or hardcoded credentials. The commands provided are standard developer tools used locally within a debugger environment.\n- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process external data from a debugger (e.g., backtraces and variable values). While this represents a potential attack surface for indirect prompt injection if the debugged application's state contains malicious strings, the skill's primary function is to assist a developer in interactive troubleshooting, which is a low-risk activity in this context.\n
- Ingestion points: User-provided LLDB output (e.g., thread info, backtraces, variable values) as described in the playbooks.\n
- Boundary markers: The skill instructs the agent to provide structured, confidence-tagged diagnoses rather than raw data reflection, which helps mitigate accidental instruction following.\n
- Capability inventory: The skill is instructional and does not include tools or permissions for automated file-writing, network access, or command execution outside the debugger context.\n
- Sanitization: No explicit sanitization of the debugger output is described, as the agent is expected to interpret the technical data provided by the user.
Audit Metadata