An intent can declare its mechanism claim and its scope conditions

Context

An intent's press release commits to an outcome, and its acceptance criteria commit to a verifiable bar — but nothing in the shape asks why the authors expect the mechanism to work, or under what conditions the claim holds. The gap shows up downstream: when a shipped intent underdelivers, the audit can compare promise to delivery but not expectation to mechanism, because the expectation was never written; and when a capability is reused outside the conditions its design assumed, nothing recorded says the assumption existed. The brief-creation interview design surfaced both: an elicited claim arrives with grounds ("we expect this to work because…") and with boundaries ("this holds while…"), and the intent shape had nowhere to put either.

Decision

The intent shape gains two optional sections, documented in the /abcd:intent surface page:

1. ## Mechanism — the mechanism claim: why the authors expect this to work, stated as a falsifiable expectation ("we expect X because Y"), not a restatement of the outcome.
2. ## Scope Conditions — the conditions under which the intent's claim holds: the population, platform, scale, or assumptions the design leans on, so a reuse outside them is a visible re-decision rather than a silent stretch.

Both sections are optional and unenforced. Whether either is ever required — for every intent, for severity: major intents, at the planning gate — is a discipline question, explicitly deferred; if enforcement comes, it arrives as its own record on the itd-84/itd-1 pattern (a rule with a staged gate), not as a side effect of this ADR.

Alternatives Considered

1. Fold mechanism into "Why This Matters". Rejected: that section argues value, not mechanism; blending them lets an intent argue importance while staying silent on why the approach should work.
2. Require the sections immediately. Rejected: a required section with no calibrated reviewer becomes boilerplate; itd-84's promotion ladder shows the order that works — shape first, discipline after evidence.
3. A separate record family for mechanism claims. Rejected: the claim is the intent's own, and adr-30's information architecture routes a feature-scoped statement to the feature's record, not a parallel store.

Consequences

• The surface page's template gains both sections with their one-line contracts; existing intents are untouched (optional means absent is valid).
• The intent audit (promise vs delivered) gains material when the sections are present: a delivered-but-mechanism-refuted outcome becomes expressible.
• The interview (itd-142) gets a committed destination for elicited grounds and boundaries at intent capture.
• Enforcement, if it ever comes, is a later, separate decision with its own evidence.