Scorecard acceptance engine

Fail-closed promotion surface for the eleven differentiable baseline-scorecard categories. Honest behind_baseline inventory is the default until claim-ledger and external-comparison evidence packages exist.

Module: scpn_quantum_control.scorecard_acceptance_engine
Composes category ids from differentiable_baseline_scorecard (does not rewrite the committed scorecard artefact).

The engine is a pure decision-support surface. Listing, validating, or asking it for a promotion decision does not mutate the canonical inventory, regenerate a scorecard, execute a benchmark, contact a provider, or certify performance.

Data layers

ScorecardCategoryRecord is the immutable inventory layer. It binds one required category to its honest status, summary, evidence labels, blockers, required evidence kinds, inventory date, and claim boundary. Construction rejects unknown categories or statuses, blank text and labels, promoted rows with open blockers, promoted rows without evidence, and behind-baseline rows without blockers.

PromoteDecision is the immutable decision layer. It records the source and requested statuses, whether the request is allowed, a reason, and any missing evidence kinds. An allowed decision cannot carry missing evidence. A decision describes the request only; it does not update a ScorecardCategoryRecord.

The registry payload is the aggregate layer. It carries the schema identifier, claim boundary, category counts, zero-blank count, and JSON-ready category rows. to_dict() converts immutable tuple fields into lists at the serialisation boundary.

Rules

Status	Meaning
behind_baseline	Honest inventory; requires blockers
at_baseline / exceeds_baseline	Only via promote with required evidence
not_comparable	Requires impossibility evidence labels

Promotion to ready statuses requires evidence identifiers covering:

• claim_ledger_promoted_row
• external_baseline_comparison

Each required kind must appear in at least one supplied identifier. The engine does not parse or authenticate the referenced package; callers remain responsible for digest, provenance, and custody validation.

Claim boundary:

scorecard acceptance engine only; behind_baseline is an honest inventory state; promote refuses invent-green at_baseline/exceeds_baseline without required evidence digests and language-safe claims

Lookup and filtering

list_scorecard_category_ids() returns all required identifiers in canonical order. get_scorecard_category() returns one immutable row and raises ValueError for blank or unknown input. iter_scorecard_categories() returns the complete immutable tuple or filters it by a supported status while preserving order.

All eleven canonical rows currently use behind_baseline and carry explicit blockers. That state is evidence-bounded inventory, not a forecast about future parity.

Promotion decisions

promote_scorecard_category() applies a fixed fail-closed sequence:

1. Validate the category and requested status.
2. Always permit keeping or returning a row to behind_baseline.
3. Require at least one evidence label before declaring not_comparable.
4. For at_baseline or exceeds_baseline, require both evidence-kind labels.
5. Reject the bounded list of unqualified promotional phrases.
6. Return an allowed engine decision while leaving artefact regeneration to a separate controlled operation.

Missing evidence produces a refused decision with missing_evidence. Banned language produces a refused decision with a bounded reason and no false claim that evidence itself was missing.

Integrity validation

assert_scorecard_acceptance_integrity() builds the canonical registry when no payload is supplied, or validates a caller-provided mapping. It requires a non-empty list of mapping-shaped rows, every required category exactly through set coverage, supported statuses, blockers for behind-baseline rows, evidence for ready rows, zero blank entries, and an exact category count.

Validation rejects malformed or incomplete payloads instead of treating them as partial success. It does not authenticate evidence identifiers or compare benchmark results; those remain separate evidence-system responsibilities.

Public API

from scpn_quantum_control.scorecard_acceptance_engine import (
    assert_scorecard_acceptance_integrity,
    build_scorecard_acceptance_registry,
    list_scorecard_category_ids,
    promote_scorecard_category,
)

reg = assert_scorecard_acceptance_integrity(build_scorecard_acceptance_registry())
assert reg["behind_baseline_count"] == 11
assert len(list_scorecard_category_ids()) == 11

# Invent-green without evidence → refuse
d = promote_scorecard_category(
    "benchmark_promotion",
    target_status="exceeds_baseline",
)
assert d.allowed is False
assert d.missing_evidence

Bounded product status

Shipped:

• Complete category catalogue and honest baseline inventory
• Fail-closed promotion decisions and integrity validation
• JSON-ready record and decision mappings
• Operator guide and real-surface tests

Outside this surface:

• Scorecard artefact regeneration or dashboard presentation
• Hermetic execution of referenced comparison packages
• Evidence authentication, benchmark execution, or provider access

Authored by Anulum Fortis & Arcane Sapience (protoscience@anulum.li)