[chamsechan]

Prune duplicate test checks while retaining coverage of failure paths and Operator contracts.

• Remove the standalone architecture asset presence gate. The existing renderer gate now owns the framework overview presence/nonempty check, with missing and empty asset regression cases.
• Remove the Catalog CLI smoke entry already covered by JSON protocol tests and its unused test-tool argument.
• Rely on CMake GoogleTest discovery instead of maintaining a second test/filter enumerator; retain metadata, disabled/list-only, skip-code and failure-propagation guards.
• Keep complete keyword response assertions. Generic Operator, SDK consumer, output pool and Node suites already cover identity, rule updates and lease reuse; the remaining Catalog snapshot test covers independence from live registries.
• Use GPT-6 Astra for Codex roles with low/medium effort for bounded work and xhigh for review/verification. No cost, latency or task-quality improvement is claimed without a comparative benchmark.

Validation: four focused CTest suites passed (quality gate contract, registration labels, diagram renderer, CLI). The canonical delivery gate passed on Linux with 467 CTest entries, zero failures, and five explicit optional-asset skips (three Inferlet model cases and two tokenizer golden-ID cases). Codex role TOML parsed successfully. Independent read-only review found no remaining merge blockers. The delivery script requires successful PR CI and successful main push CI for the exact merge commit.

Review corrections: restored the unique quality-gate negative tests and moved the overview asset check into the retained renderer gate before delivery. No production algorithm or public SDK contract changed.