adaptive-reasoning
Pass
Audited by Gen Agent Trust Hub on Mar 4, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: No malicious patterns, obfuscation, or data exfiltration attempts were detected. The skill is composed of purely instructional markdown and logic for internal state management.
- [PROMPT_INJECTION]: The skill provides logic for the AI to interpret user requests and adjust its internal behavior (activating reasoning mode). While these are directives for the AI, they are designed to improve response quality and do not attempt to bypass safety guidelines, extract system prompts, or override fundamental safety constraints.
- [COMMAND_EXECUTION]: The skill mentions commands like
/reasoning onand/status. These are internal agent control commands or tool calls (session_status) used to manage the model's internal processing state, not shell commands or system-level executions.
Audit Metadata