[ajhcs]

The existing model view returned native records separately from scoped concept definitions and terminology bindings. This adds an opt-in assembled view that presents those meanings with their explicit dependencies and source provenance, while preserving the existing view invocation.

• Join existing foundation, English and workflow records with linked native declarations. Retain roles, relation direction, allocation exclusions, conditions, units, authored clocks and evidence stages. Report omitted or unresolved dependencies, including partially known native connector contexts.
• Parse captured model snapshots, preserve original diagnostic locations, reject input drift and enforce the byte cap on the actual emitted JSON.
• Connect one shared drafting/output-review guide to all eleven skills. Add a 37-case synthetic corpus with distinct authored interpretive expectations and deterministic controls.

Three independent subagents reviewed the implementation. Their findings received regression tests and fixes, followed by independent reruns of the original failure cases. Unrestricted prose/code interpretation remains an assessment; generation-quality gains have not been benchmarked.

Validation: the full regression passed 71 suites / 1,156 tests with zero skips, using the pinned native parser and symbolic dependencies. Source bytes remained stable throughout the run. Official gallery/showcase rebuilds and renders reproduced all 66 non-manifest files, including nine PNGs; only the two engine-fingerprint manifests changed. The original dirty checkout and its staging were preserved.