mode-switcher

Pass

Audited by Gen Agent Trust Hub on Apr 8, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The 'night', 'yolo', and 'unhinged' modes instruct the agent to operate with 'autoAccept' permissions and skip user confirmation prompts, which suppresses standard safety loops and user oversight.
  • [PROMPT_INJECTION]: The skill facilitates a high-autonomy environment without safeguards against indirect prompt injection in processed data. 1. Ingestion points: Project files and task lists (e.g., tasks/todo.md). 2. Boundary markers: Absent. 3. Capability inventory: High-autonomy file-write operations and extensive tools across all CCC domains (DevOps, SaaS, etc.) as specified in unhinged.md. 4. Sanitization: Absent.
  • [COMMAND_EXECUTION]: The 'unhinged' mode encourages 'aggressive refactoring' and architectural rewrites without user approval, creating a risk of autonomous destructive operations if the agent follows malicious instructions from the processed codebase.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 8, 2026, 02:36 PM
Security Audit — agent-trust-hub — mode-switcher