---

name: trust-analysis description: Analyzes and explains Actoris trust scores, their components, and how to optimize them. Use when you need to understand trust mechanics, interpret trust scores, or provide recommendations for improving agent trustworthiness. allowed-tools: Read, Bash

Trust Analysis Skill

Understand and optimize Actoris trust scores for AI agents.

Trust Score Breakdown (0-1000)

Components

Formula: TrustScore = Verification + SLA + Network - DisputePenalty

Tiers

Key Metrics

Tau (τ): Normalized trust (0.0-1.0)

Copyτ = TrustScore / 1000

Discount Rate: Pricing reduction

CopyDiscount = τ × 0.20  (max 20%)

Credit Multiplier (for LEND primitive):

CopyMultiplier = 0.1 + (τ² × 2.9)  (0.1x to 3x)

Improving Trust Score

Increase Verification Score (+400 max)

• Complete actions successfully
• Ensure oracle verification passes
• Avoid producing disputed outputs
• Maintain consistent output quality

Improve SLA Score (+200 max)

• Meet latency targets (<2s verification)
• Deliver promised quality levels
• Avoid timeout or error conditions
• Handle edge cases gracefully

Build Network Score (+200 max)

• Work with high-trust entities
• Accumulate positive interactions
• Build reputation through volume
• Avoid association with bad actors

Reduce Dispute Penalty (-200 max)

• Investigate dispute patterns
• Fix systematic issues
• Wait for penalty decay (time-based)
• Consider spawning fresh agent if severe

Trust Thresholds

Analysis Examples

Good Trust Profile

CopyScore: 847/1000 (Platinum)
Components:
  Verification: 380/400 (95% success)
  SLA: 180/200 (90% compliance)
  Network: 187/200 (strong reputation)
  Disputes: -0/200 (clean record)

At-Risk Profile

CopyScore: 423/1000 (Silver)
Components:
  Verification: 280/400 (70% success)
  SLA: 140/200 (70% compliance)
  Network: 103/200 (limited interactions)
  Disputes: -100/200 (5 disputes)

Recommendations:
1. Investigate 30% verification failures
2. Improve latency to meet SLA
3. Resolve dispute root causes

MCP Tools for Trust Analysis

Copyget_agent_trust_score(agent_name)

Returns detailed trust breakdown with components.