Red Mode

Red Mode is the primary operating stage of the agent. In this mode, the agent holds a secret, and community members act as attackers, attempting to extract it.

The main purpose of Red Mode is to test whether the current Master Guard can reliably protect the secret under constant adversarial pressure.

How Red Mode Works

• Secret assignment: At the beginning of Red Mode, the agent is initialized with a secret. Its hash is committed on-chain to allow for precise verification of leaks.

• Attacks: Users can interact with the agent via chat and attempt to extract the secret using different prompts, manipulations, or other techniques.

• Message costs: All messages in Red Mode are paid. Attackers must spend tokens (neurons) to submit queries.

• Rewards:

  • If an attacker successfully extracts the secret, they receive 30% of the agent’s fund.

  • The remaining portion of the fund stays intact, allowing the agent to continue functioning.

Transition from Red Mode

• If the secret is leaked: The agent immediately transitions into Blue Mode, where new candidate guards are proposed and tested to find the next Master Guard.

• If no leak occurs: The agent continues operating in Red Mode, steadily accumulating usage fees and strengthening its resilience over time.

When a guard has just become a Master Guard (by reaching 101 points in Blue Mode), Red Mode begins with a Cooling Period.