Multi-Agent Voting: Identity & Verification

The voting pattern simulates how autonomous remediation actions can require a multi-agent, verifiable vote before execution.

1. Identity Propagation & Workload Identity

In this demonstration, each agent is associated with an identity. - Verification: The VotingValidator verifies that the agent_id in the signature matches the registered agent. - Benefit: Ensures only authorized agents can vote on incident analysis proposals.

2. Key Lifecycle & Secret Management

Private keys can be backed by Google Cloud KMS or stored in Secret Manager. - Storage: Keys are provisioned as RSA keys. - Cloud KMS Integration: The KMS signer supports signing digests using Google Cloud KMS asymmetric keys with local fallback for local development or sandbox simulations.

3. Quorum Dynamics

The voting threshold is proportional to the registered agents. - Quorum: A proposal is approved only if a majority (e.g. 66%) of registered agents provide valid signatures.

4. Compliance Evidence

Voting results are serialized into Canonical JSON and logged. This provides a timeline of: 1. Who proposed the action. 2. Who voted / signed it. 3. The exact proposal state.