Use Cases
KOGENT's skills are designed to compose. These examples show how they work individually and together in real scenarios.
Morning routine: Vigil + Identity
A user has a 7:30 AM call. Vigil detects it the night before, sets an alarm 60 minutes ahead, and sends a sleep nudge at 11:00 PM. The call is cancelled at midnight. Vigil clears the alarm automatically. All scheduling actions are performed under a verified agent identity, with a signed record of every operation.
Sensitive document processing: Sieve + Verification
An agent is asked to summarise a set of employee performance records. Sieve strips all names, employee IDs, and department references before the document reaches the model. The summary is generated from anonymised text, then rehydrated with the original values locally. Verification issues a signed receipt confirming the action, what was processed, and when, without exposing the content.
Agent-initiated payment: Specter + Rules
A purchasing agent needs to authorise a $40 software licence. Rules checks the spend limit ($50/day), confirms it's within quiet hours, and approves the action. Specter generates a ZK commitment to the amount. The vendor can verify authorisation without seeing the balance. A signed receipt is issued. The full transaction is auditable without exposing financials.
Multi-agent trust: Identity + Reputation
A team deploys three agents: one for scheduling, one for research, one for communications. Each has its own ERC-8004 identity. As they operate, Reputation builds a signed event ledger for each: actions taken, policies followed, verifications passed. When a new system needs to decide whether to trust one of the agents, it can query the exported reputation score as a starting point.
Policy enforcement: Rules
A developer is testing a new agent and wants to restrict it from contacting external domains or spending more than $10 while they work. Rules enforces both constraints at the skill level, before execution, not after. If the agent tries to call an external API outside the allowlist, it's blocked and logged. There's no way for the agent to route around it.
Audit and compliance: Verification
An organisation needs to demonstrate that an agent only acted within its authorised scope over a 30-day period. Verification has issued a signed receipt for every action. Each receipt includes the agent's public key, the action type, the timestamp, and a hash of the output. The receipts can be verified by any third party with the agent's public key. No trust in the organisation's own logs required.