[ancongui]

Studio presents Codex MCP calls with a clear tool name, provider-reported status and bounded output, while keeping native runner receipts and product verification separate. Recovery replies label previous-attempt verification before its totals so queued work does not appear to have passed historical checks again.

The 0.2.30 preparation refreshes the controlled activity guide and retains explicit full-event loading, archive integrity checks and existing delivery identities. It includes the supported asyncio launch boundary and isolated-exporter correction from #41.

Validation:

• 125 affected host tests passed; one native database test was explicitly excluded. Ruff, formatting, all 229 module compilation checks and strict types passed.
• Offline wheel and source archives passed exact source, metadata, font and Studio payload checks. An isolated installation passed MCP projection, singleton tool approval, runtime admission and recovery-feedback checks.
• Strict documentation built 60 pages and social cards; 19 documentation checks passed. The unchanged frontend retains its qualified 549-unit, 94-documentation-browser and five Pages checks.
• All 488 controlled screenshots, capture inputs and frontend assets match the reviewed 0.2.30 preparation. The preserved capture evidence includes the original 67-pass/one-timeout run and the exact failed case's successful production-preview successor.

Provider messages and controlled display fixtures do not establish fresh-run Codex tool access, a completed CRM or generated-product acceptance. This change adds no execution budget or recovery authority.

Fresh full CI for 6a0fa048 passed: 3,737 Python tests (51 skipped, four warnings), 549 Studio units, 735 browser cases and 94 documentation browser cases. The downloadable CI distributions match the reviewed source, Studio, metadata and license payloads, and the qualified isolated installation. Release 0.2.29 is published and verified. Normal merge and 0.2.30 tagging have final review approval.