roblox-collaboration-mode
Pass
Audited by Gen Agent Trust Hub on Oct 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill consists entirely of instructional markdown that defines behavioral logic for the agent. It does not include any executable scripts, tool definitions, or network-based dependencies.\n- [INDIRECT_PROMPT_INJECTION]: The skill logic involves parsing user input to determine whether to operate in Peer or Autonomous mode, which creates a potential surface for indirect influence.\n
- Ingestion points: Processes natural language from user requests (e.g., "build X", "let's redesign Y") to switch between behavior modes.\n
- Boundary markers: No specific delimiters or markers are defined to isolate user data from the agent's logic.\n
- Capability inventory: The skill itself requests no tool access and lacks the ability to perform file system or network operations independently.\n
- Sanitization: The skill explicitly instructs the agent that user requests to "accept risk" cannot override mandatory platform policies, providing a safety-focused constraint on its autonomy.
Audit Metadata