Skip to content

Latest commit

 

History

History
119 lines (70 loc) · 2.31 KB

File metadata and controls

119 lines (70 loc) · 2.31 KB

Guardian Decision Artifact — v0 (Semantic Invariant)

Status

Draft — Semantic Proposal (Non-binding)


1. Problem

In agent systems, execution receipts are often treated as primary evidence.

However, receipts alone do not establish whether an action was authorized — only that it occurred.

This creates ambiguity for:

  • audit
  • replay
  • cross-engine verification

2. Core Invariant

A Decision Artifact is a first-class, independently verifiable object that represents the outcome of a policy evaluation.

It MUST NOT be:

  • reconstructed from execution receipts
  • reduced to metadata attached to execution logs

3. Separation of Concerns

Layer Responsibility
Decision why an action is allowed
Execution what actually happened
Receipt record of execution outcome

Invariant: Decision ≠ Receipt


4. Verifier Semantics

A verifier SHOULD:

  1. Validate the integrity and signature of the decision artifact
  2. Evaluate the policy reference and evaluation method
  3. Determine replayability based on evaluation type

5. Deterministic Boundary

Replayability is only guaranteed when:

  • the evaluation method is deterministic
  • inputs are fully specified

This corresponds to Proof of Computation (PoC).

Probabilistic outputs MAY be signed, but:

  • are not replay-equivalent evidence
  • require additional interpretation

6. Minimal Fields (Abstract)

A decision artifact SHOULD minimally include:

  • intent_id
  • policy_reference
  • evaluation_method
  • decision_outcome
  • decision_hash

This set is intentionally minimal and engine-agnostic.


7. Cross-Engine Implication

In cross-engine scenarios:

  • execution traces may not be available
  • only signed artifacts can be exchanged

Therefore:

Verification MUST rely on decision artifacts, not execution receipts or execution trace reconstruction.


8. Non-Goals

This document does NOT define:

  • execution formats
  • transport protocols
  • runtime enforcement mechanisms

It is purely a semantic invariant layer.


9. Summary

A system is verifiable across engines only if:

  • decisions are independently representable
  • verification does not depend on execution reconstruction

This document defines that boundary.