⌛🎯👈☿📎🏺 AXN:062E.GOVERNANCEClaim Status Packet v0.6OBJECT: DRAFT · UNIMPLEMENTEDspxi.dev/cspCSP v0.6 · #1529 · AXN:062E.GOVERNANCE
← spxi.dev SPXI-TLP v2.2 · ASSEMBLY-RATIFIED · 2026-05-25

The Claim Status Packet

CSP v0.6 · a schema so that a claim's state travels as data · Lee Sharks · CC BY-SA 4.0 · deposit #1529

UNIMPLEMENTED. No generator, no checker, no anchors, no generated packet. The specification states throughout that its own bindings would fail its own tests today. It is deposited as a schema, not as a deployed instrument.

The defect it addresses

Two hand-maintained copies of one fact will diverge. The only question is when.

The archive ruled this once already: deposit counts hand-set across the fleet drifted to 57% of true, and the repair was a binding rather than more careful editing. The same class recurred inside a specification batch — a document's version diverged from its own machine-readable layer three times, and a single status string was hand-maintained in six locations. Corrections logged as sweep residue were largely those copies disagreeing. Recorded as editing failures, they were a schema failure.

Five objects

objectjob
claim + statestable IDs, typed state, document and version facts
relationssupports · interprets · qualifies · generalizes · rebuts · supersedes
provenancewhy, how and by whom each transition happened
bindingswhere each state renders, with anchors and hashes
policyinvariants and legal transitions

The rule

State is typed and claim-local. Relations do not copy state. They license specified state transitions on specified dimensions. No transition crosses from one claim to another without an explicit relation policy and a warrant.

Most relations license nothing. motivates and interprets — the two commonest edges — transmit no state at all. That is what prevents a candidate interpretation from attenuating the observation it interprets: not because claim kinds are sealed, but because the relation transmits nothing.

Generate facts, lint argument

Mechanical facts — version, dates, status enums, inclusion flags — are projected from the packet at build time. Claims, interpretations and scope-sensitive sentences remain authored.

An earlier draft got this wrong and the specification records it. It specified a checker over hand-maintained copies while describing itself as a single-source binding. That improved six unaudited copies to six copies plus a linter — which is not the repair.

The metric

unsupported_attenuation — not raw weakening. Early drafts overclaim and editorial narrowing is the work, so a healthy pipeline weakens often. What matters is weakening with no evidence, scope or proposition change on the claim, and no relation licensing propagation from whatever triggered it.

Undefined where no weakening occurred — a document with no attenuation has demonstrated nothing about attenuation.

Named vulnerabilities

Layer smuggling — a claim moved to a lower kind acquires protection it has not earned, and the schema then enforces a wrong state more reliably than prose could. Undeclared surfaces — unchecked, drifting freely while the report reads clean. Cleanliness relocating the mess rather than removing it. False harmony — contradictory states held indefinitely; preservation is not resolution. Claim-identity smuggling — a materially new proposition inheriting the old one's provenance by keeping its ID.

What it does not do

It does not make claims correct. It makes state consistent and transitions legal. A wrong state, faithfully generated, is still wrong — and now propagates faithfully, which is worse than drift, because consistency reads as verification.

And the deepest limit is not in the schema. Data drifts; practice does not. The generator only runs if someone runs it. A bet, not a guarantee.

Related: Four Interfaces of Provenance Transformation #1528, which states the status law this schema would make enforceable · The Lucente Extension #1527, which the first worked packet targets.