Truth Machine / Kit

Exact Review

Bind review to the exact proposed content by digest, so what was accepted is provably what was applied.

Tier: 3 · Normative weight: implements TM-CHG-001 Requires: Change Packet

Problem it removes

Approval attaches to a description — a summary, a conversation, a title — while the applied content is whatever the working tree held at apply time. Between review and application the proposal can drift, and no one can later prove it did not.

Mechanism

Review binds an identified, exact proposal: the reviewer records the digest of the precise content reviewed — in the Git-native shape, a canonical serialization of the packet, or the staged diff — and acceptance stores that digest in the change packet (proposal_sha256) together with the reviewer and an acceptance evidence packet. At application, the digest of what is being applied must equal what was accepted; any substitution is detectable rather than trusted away (TM-CHG-001). Expected-state identity travels with the proposal, so review is also bound to the state it reasoned against.

Mechanisms vary — content-addressed proposals, signed envelopes, a Git commit whose parent is verified — and the requirement is on the behavior, not the mechanism: unreviewed substitution must be detectable.

Cost

Authoring: a digest command and where to record it. Per change: one hash at review and one comparison at apply. Per session: none.

Do not adopt when

There is no reviewer separate from the applier, and no gap between acceptance and application in which substitution could occur — though even a single operator gains a self-audit trail from it.

Evidence

Assessed in kinra-corpus and in the reference instance’s own frozen standard, which reviews staged binary diffs by digest. Project proposals evidences exact-source review of proposal packets.

Template

None — the fields live in the change packet template; the digest discipline is a convention.

Atomic Transition enforces the matching expected state at apply time.