Foundation · TEST_ONLY

CONTRACT REVIEW · TEST_ONLYReview-candidate contract and local fixture evidence only · no distributed SDK, operational API, publisher parity, or decision authority

Separate what was known then from what happened later.

Why a Decision Episode preserves the original context and records later outcomes in a separate layer.

01

An Episode freezes the context available at decision time

To prevent hindsight from rewriting the past, bind the case snapshot, selected Option, open assumptions, and evidence references to one temporal boundary.

  • DecisionCase and version
  • Recorded Decision and human actor
  • Opaque references current at the time
  • Open dissent and evidence requests
  • TEST_ONLY assurance and authority NONE
02

Record later outcomes separately

A later good or bad outcome does not retroactively change the authority or evidence available at decision time. Add an evaluation record that references the original episode digest.

  1. 01
    Seal the original snapshot

    Fix DecisionCase and Decision Episode digests first.

  2. 02
    Receive the outcome separately

    Do not overwrite the prior record with a new HARBINGER or outcome reference.

  3. 03
    State the evaluation rule

    Record success or failure criteria and the evaluation time.

  4. 04
    Keep learning separate from authority

    Learning may inform a later case but cannot mutate policy or create execution authority.

03

The current implementation validates structure only

Decision Episode packages are local TEST_ONLY structural contracts and fixture evidence. They do not claim publisher parity, independent verification, an operational store, or production readiness.