ultrathink
Pass
Audited by Gen Agent Trust Hub on Sep 21, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill acts as a cognitive framework for managing reasoning depth. It does not contain any malicious code, command execution, or data exfiltration logic.
- [SAFE]: It implements defensive patterns by requiring risk assessments and user confirmation for 'T2+' risk-level actions and 'Mode 4' decisions.
- [SAFE]: Interaction with the local file system (writing decision memos to
~/.claude/kb/outputs/) is used appropriately for knowledge management and record-keeping within the agent's environment. - [SAFE]: The skill references standard tools (e.g.,
mcp__sequential__sequentialthinking,claude-mem) and related internal skills without installing external or untrusted dependencies.
Audit Metadata