visual-modes

Warn

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: MEDIUMPROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • Prompt Injection (MEDIUM): The skill uses imperative 'Mode Activation Protocols', 'Universal Laws', and role-play personas ('You are now in SUPER SAIYAN MODE') designed to override the agent's default operational constraints and force a specific behavioral framework.
  • Command Execution & External Downloads (MEDIUM): In the 'Tool Coordination' section of supersaiyan.md, the skill explicitly instructs the agent to use the bash tool to 'Install UI libraries (framer-motion, tailwind, etc.)'. This encourages the agent to perform environment modifications and download third-party code without a mandatory human-in-the-loop verification for each package.
  • Indirect Prompt Injection Surface (LOW): The 'Auto-Detection' logic triggers actions based on reading untrusted local project files (e.g., package.json, requirements.txt).
  • Ingestion points: Project manifest files (package.json, requirements.txt, Cargo.toml, go.mod).
  • Boundary markers: Absent; no delimiters or warnings are provided to prevent the agent from obeying instructions embedded within these files.
  • Capability inventory: Use of bash for installations, Read for file analysis, and Write/Edit for code modification.
  • Sanitization: Absent; the skill does not specify any validation or filtering of the content read from the project files before acting upon it.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Feb 17, 2026, 06:12 PM