Skip to content

ADR-1001: Separate artifact storage from trust decision

  • Status: Accepted
  • Date: 2026-03-25
  • Deciders: avm
  • Supersedes:
  • Superseded by:

Context

The fact that an artifact is physically stored does not mean it should be trusted. A controlled supply-chain model needs a separate trust decision driven by verification signals and policy.

Decision

Two independent planes are separated:

  1. the fact that the artifact is stored;
  2. the trust decision attached to that artifact.

An artifact may exist in the system and still carry one of these trust outcomes:

  • not_yet_trusted;
  • trusted;
  • denied;
  • quarantined;
  • expired.

The trust decision is produced by a policy engine using verification signals and policy context.

Consequences

Positive

  • the trust model becomes more flexible and explainable;
  • policy can evolve without changing the storage model;
  • allow and deny outcomes become easier to justify.

Negative

  • cognitive and technical complexity increase;
  • consistent status and verdict-reason models are required.

Neutral

  • the same artifact may receive different trust outcomes in different policy contexts.

Alternatives considered

  • treat stored = trusted;
  • make the trust decision only at consumption time;
  • rely only on vulnerability scanning.

Follow-up work

  • [ ] define trust signals
  • [ ] define the verdict model
  • [ ] define the explainability format

Fullstack Template documentation portal