> ## Documentation Index
> Fetch the complete documentation index at: https://docs.synclear.io/llms.txt
> Use this file to discover all available pages before exploring further.

# Trust Scoring Model for AI Agents in Synclear Protocol

> How Synclear calculates agent trust scores from 0 to 1000, which behaviors move scores up or down, and how score tiers govern agent spending capabilities.

The Synclear trust scoring system provides a continuous, on-chain measure of an AI agent's operational reliability and protocol compliance. Rather than treating all agents identically regardless of their track record, the protocol uses the trust score to calibrate the operational latitude extended to each agent — rewarding disciplined behavior with expanded capabilities and applying conservative constraints to agents with limited or negative history.

Trust scores are maintained by the `AgentScorekeeper` contract, which processes on-chain events from the `MandateRegistry` and related protocol contracts to update scores in real time. The score is a public, queryable value associated with each registered agent address.

## Score Range and Starting Point

Trust scores are integers on a scale from **0 to 1000**. All newly registered agents begin at a score of **250**, placing them in the Standard tier from day one. This baseline reflects that a new agent has no negative history but also no demonstrated track record. The 250 starting point gives new agents meaningful operational capacity while the protocol gathers evidence of their behavior.

## Score Tier Table

An agent's score tier determines the maximum daily volume the `MandateRegistry` will permit, independent of what the agent's mandate nominally authorizes. The effective limit is always the lower of the mandate-defined limit and the tier ceiling.

| Score Range | Status       | Max Daily Volume |
| ----------- | ------------ | ---------------- |
| 0 – 200     | Probationary | \$10,000         |
| 201 – 500   | Standard     | \$100,000        |
| 501 – 800   | Trusted      | \$500,000        |
| 801 – 1000  | Premium      | Negotiated       |

Premium-tier agents work directly with Synclear to establish custom volume limits appropriate to their operational profile and the principal's institutional relationship with the protocol.

## Factors That Increase Score

The `AgentScorekeeper` applies positive scoring increments for the following observable behaviors:

<Accordion title="Consistent Mandate Compliance">
  Every transaction that passes all mandate validation checks and executes successfully contributes a small positive increment to the agent's score. The increment is weighted by transaction size — larger compliant transactions carry slightly more scoring weight than small ones, reflecting the higher stakes involved. Compliance increments accumulate over time and are the primary mechanism through which agents advance through tiers.
</Accordion>

<Accordion title="Successful Obligation Management">
  When an agent executes `REPAY` actions that service outstanding credit obligations on time or ahead of schedule, the `AgentScorekeeper` records a material positive event. Consistent on-time repayment history is one of the strongest positive signals in the scoring model, reflecting the protocol's credit-native identity.
</Accordion>

<Accordion title="Tenure and Track Record Length">
  Agents that have been active for longer periods with clean histories receive a gradually increasing tenure bonus applied to each positive event. This reflects the protocol's confidence that a sustained clean record is more predictive of future behavior than a short burst of compliant activity. The tenure bonus begins accumulating after 30 days of active operation.
</Accordion>

<Accordion title="Absence of Adverse Events">
  Agents that maintain extended periods without any negative scoring events receive periodic stability bonuses. These are applied at 30-day intervals and are proportional to the agent's current score — agents closer to tier boundaries receive larger bonuses, accelerating the transition to the next tier for agents with consistently clean records.
</Accordion>

## Factors That Decrease Score

<Accordion title="Mandate Violations">
  Any attempt by an agent to submit a transaction that is rejected by the `MandateRegistry` for a policy violation — wrong action type, disallowed asset, exceeded limit — generates a negative scoring event. The severity of the decrement scales with the nature of the violation: an exceeded per-transaction cap generates a smaller decrement than an attempt to use an action type not listed in the mandate. Repeated violations within a short window trigger an accelerating penalty curve.
</Accordion>

<Accordion title="Liquidation Events">
  If an agent's positions are subject to protocol liquidation due to collateral ratio deterioration, the `AgentScorekeeper` records a significant negative event. Liquidations indicate that the agent failed to manage risk within its mandate's operational scope, either through inaction or through position-taking that left collateral dangerously exposed.
</Accordion>

<Accordion title="Suspicious Activity Flags">
  The Synclear risk engine monitors for anomalous agent behavior patterns — unusually rapid transaction sequences, systematic attempts to probe mandate boundaries, transaction patterns inconsistent with the agent's registered purpose, or interactions flagged by integrated AML/compliance screening. Confirmed suspicious activity flags generate large negative scoring events and may trigger automatic mandate suspension pending principal review.
</Accordion>

<Accordion title="Overdue Obligations">
  Credit obligations that pass their scheduled service date without a `REPAY` action generate an overdue penalty that increases in magnitude each day the obligation remains unserviced. The scoring impact of overdue obligations is designed to exceed the cumulative positive increments from normal operation, making timely repayment the most score-sensitive single behavior.
</Accordion>

## Score Recovery

Score decrements are not permanent penalties — they are events that the score model must overcome through subsequent positive behavior. An agent that experiences a liquidation event can recover to its pre-event score through consistent compliant operation, though the recovery period is intentionally calibrated to be meaningful (typically several weeks of active operation for a significant adverse event).

Scores cannot be manually adjusted or reset by principals. The integrity of the scoring system depends on it being exclusively driven by on-chain protocol events.

<Tip>
  The fastest path to a higher trust score is prioritizing `REPAY` actions and maintaining consistent, moderate-volume operation rather than attempting to maximize volume immediately. Agents that demonstrate reliable obligation management advance through score tiers significantly faster than agents with higher raw volume but inconsistent repayment behavior. Configure your agent to service obligations proactively — ahead of due dates where possible — and avoid transaction patterns that trigger rate limit or cooldown rejections, as each rejection generates a small but compounding negative scoring event.
</Tip>

## Querying an Agent's Score

```js theme={null}
const scoreData = await agentScorekeeper.getScore(agentAddress);
// returns {
//   score: number,           // current score (0-1000)
//   tier: string,            // "PROBATIONARY" | "STANDARD" | "TRUSTED" | "PREMIUM"
//   lastUpdated: number,     // Unix timestamp of most recent scoring event
//   effectiveDailyLimit: BigInt  // active daily volume cap based on current tier
// }
```