Skip to content

Latest commit

 

History

History
28 lines (18 loc) · 957 Bytes

File metadata and controls

28 lines (18 loc) · 957 Bytes

AgentTerm Integration

AgentTerm is an operator surface. It should consume Agent Machine facts and decisions rather than owning machine-runtime policy itself.

Current bootstrap command surfaces

Agent Machine exposes:

  • agent-machine doctor;
  • agent-machine probe;
  • agent-machine render ...;
  • agent-machine activate evaluate ....

These commands can feed terminal-native operator workflows and ChatOps summaries.

Required future integration

AgentTerm integration should provide:

  • command palettes for doctor/probe/render/activation evaluation;
  • readable summaries of ActivationDecision and fail-closed reasons;
  • links to receipts and evidence artifacts;
  • clear distinction between dry-run evaluation and real activation;
  • no display of secret values or raw cache contents.

Safety rule

AgentTerm must not convert a render/evaluation artifact into runtime activation without Policy Fabric admission and Agent Registry grants.