A03 — Finite primitives composed into procedural structure

• Input [inference]: A procedural artifact whose reasoning steps can be described independently of domain labels, consistent with the definition of "Primitive Reasoning Operation" as "a provisional low-level reasoning function used to describe procedural structure independently from domain labels" (theory doc, Core Definitions).
• Operation [direct]: Identify primitive reasoning operations and compose them into a "Procedural Reasoning Grammar," defined as "a hypothesized composition of reasoning primitives that structures analysis, diagnosis, evaluation, choice, verification, reflection, and reassessment" (theory doc, Core Definitions).
• Output [inference]: A primitive-level representation of an artifact's reasoning structure, sufficient to attempt reconstruction or synthesis of procedural structure (inferred from Observations 3 and 4's discussion of "reconstruct" and "synthesize").
• Scope conditions [direct]: Reasoning-heavy procedural artifacts; the theory doc's fourth provisional claim states "primitive sufficiency appears weaker in coordination-heavy problems than in reasoning-heavy problems."
• Failure conditions [direct]: The falsification criteria state the theory weakens if "independent extractors do not converge," "primitive vocabulary keeps expanding," or "reconstruction fails outside familiar domains."
• Predicted effect [direct]: "Independent extractors will converge more strongly on reasoning-heavy artifacts than on coordination-heavy artifacts" and "primitive vocabulary growth will slow for reasoning artifacts before it slows for coordination artifacts" (Predictions).
• Measurement [inference]: Vocabulary growth and extractor convergence are implied as the measured quantities by the predictions above, though no scoring procedure is defined in the permitted set.
• Implementation test [absent]: The extraction, normalization, reconstruction, and holdout-testing procedure described in the v1.1 lead card is not present in this role's permitted canonical set and is therefore not asserted here.
• Decision rights [absent]: Not addressed.
• Uncertainty handling [inference]: Observation 3 notes vocabulary stabilization was not complete and reconstruction results were mixed ("strong or moderate... for most artifacts"), implying uncertainty about vocabulary status should be preserved rather than treated as settled; this is an inference from the hedged observation wording.
• Verification [direct]: Next Required Work calls to "Re-run FE-012A with independent extractors and audited merge disagreements" and "Re-run FE-012B with independent reviewers and a larger set of problems not previously used."
• Provenance [absent]: Not addressed in the permitted canonical set.
• Implementation cost [absent]: Not addressed.
• Limitations [direct]: Observation 3 states the internal run "showed [incomplete] vocabulary stabilization" and "remains vulnerable to same-model bias and extractor non-independence." Observation 4 states coordination-heavy problems "exposed missing primitive pressure" against the current frozen vocabulary. Open Questions ask whether Compare and Prioritize are durable primitives or temporary repair terms, and whether the theory can "survive independent human testing rather than internal same-model runs."

Statement basis

• Direct: Definitions, scope claim, failure/falsification conditions, predicted effects, and verification requirement.
• Inference: Input/output framing and the uncertainty-preservation implication of Observation 3's hedged wording.
• Proposal-only: None identified within the permitted canonical set for this candidate.
• Absent: Extraction/normalization procedure, decision rights, provenance rule, implementation cost, and any named matched baseline.