Discovery acceptance and limits

Authority: DEC-0024 (repository reference). Status: accepted through DEC-0024 after independent classification and normative promotion.

Exact payload

The normative rule payload is README.md from the UTF-8 heading ## Normative boundary to immediately before ## Evidence and deferred work, including intervening line breaks. It contains DE01-DE08. Introductions, the selection rationale, deferred-work prose and broader draft contracts are informative only.

Adoption covers inputs.json and expectations.json in the separate discovery-evidence-v0.1 corpus. Each case's inlined source material, authored premises, declared review scope and paired expectation establish only its bounded conformance finding. Equivalent decisive reasons are acceptable. DQ13 covers full context disclosure; DQ20 lacks a candidate and is not_assessable; all other cases are restricted. The 20 questions contain 39 candidate assessments, not 39 independent economic findings or source-extraction trials.

The JSON structure is editorial fixture notation. No canonical fields, schema, grammar, API or implemented classifier are adopted. The original AQ packet, SC01-SC12, DEC-0022 and accepted core/research corpus remain unchanged. Draft DS/IF/CH/FC clauses and DEC-0018 to DEC-0021 remain proposed. No new economic operation is authorized.

Byte identity

Hashes identify bytes, not publisher authenticity, semantic truth or review independence. source-pins.json identifies the unchanged research, proposal and accepted-rule sources.

Review and promotion source material

Two fresh read-only classifiers received isolated inputs and applicable rules at the initial frozen candidate a257a8bac129b023d8e0f8d6f40d602f1c918d77, with the oracle withheld. Both found DQ02's misapplied operation label and DQ20's scope metadata inconsistency. Both missed DQ13's query-interpretation review-state and tie-policy omissions; the separate content reviewer returned that full-context candidate for correction.

The corrected payload was frozen at 6aab9a00a890055d9c32c06117dfc3028d1f5ce4. Both classifiers reassessed the three affected cases, verified 17 cases unchanged and reaffirmed their 34 remaining findings. Each final report records 19 meets, 19 violates and one not_assessable, with separate operation outcomes and decisive reasons. These are informed correction reviews. The content reviewer accepted the corrections. All original reports and correction acknowledgments remain in review-results.json. Root links the reports to the separate oracle; classifiers never inspected that oracle.

The protocol excluded parent dossiers and live enrichment. Reused questions remain development source material; agreement does not establish independent extraction accuracy.

Two read-only promotion reviewers accepted candidate fa1bcee68e1d239d97109eaddb3584205b27372e after inspecting exact pins, case coverage, exclusions, source material fidelity and both full classification records. One performed prior content review and the other prior classification; both promotions are informed. Root records the acceptance disposition under the user's product-owner delegation. Final clean-head audit, allowed-path census, local checks and reviewed PR integration remain required; verify the integrated SHA.

Fixture tests check hashes, projection fidelity, scope and review linkage. They do not classify economic records, certify source interpretations, demonstrate interoperability or establish extractor accuracy. Shared platform errors and prior source familiarity remain possible. No new human validation is claimed.