Skip to content
This repository was archived by the owner on Apr 12, 2026. It is now read-only.

feat: trust scoring — dynamic agent reputation from success/failure history #1524

@jpleva91

Description

@jpleva91

Context

Harvested from Microsoft AGT AgentProfile + TrustTracker. Agents that succeed more get higher trust scores, unlocking more capabilities. Agents that fail get restricted.

Proposal

  • Track agent success/failure events per session
  • Compute trust score (0-1000)
  • Gate capabilities on trust level (new agent = restricted, proven agent = full access)
  • Feed scores to Sentinel for cross-session analysis

Harvest source

Microsoft AGT trust_policy.py — TrustCondition + TrustTracker pattern

Metadata

Metadata

Assignees

No one assigned

    Labels

    agent:claimedAgent dispatched — do not re-dispatchenhancementNew feature or requestsprintCurrent sprint priority

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions