[tianxiaochannel-oss88]

Summary

The normal conversation footer had separate model, reasoning, summary, authorization, and sandbox controls. It now exposes three approval shortcuts beside the input—Request approval, Help me approve, and Full access—and combines model selection and reasoning strength in one popover. The slider fill follows the rounded thumb.

Independent authorization and sandbox defaults remain in Settings. Request approval and Help me approve preserve the sandbox. Full access changes both axes only after explicit confirmation; cancel preserves both, and switching back restores the previous sandbox (or the workspace default after reopening). Inherited always authorization with a restricted sandbox is displayed accurately. Confirmation state stays in the compact button's accessible name and tooltip.

Validation

• GitHub full CI and CodeQL passed on 59735aa76d4e8ba94a51f82da4bf835d83d0fe99: all 15 checks passed, including the required aggregate, all four Daemon shards, Linux contracts, and the real macOS helper/Seatbelt smoke. CI run: https://github.com/ling-stack-org/ling-stack/actions/runs/39954759834
• Web build/typecheck, lint, and i18n audit passed; Web unit tests: 575 passed.
• Composer, request payload, model/default restore, approval confirm/cancel, sandbox restoration, and conversation regressions: 21 passed.
• CI Web QA cases covering Chinese compact approval controls, full-access evidence/focus, built-in skill metadata, and Chinese QA member identity: 6 passed.
• Canonical normal-composer layout/focus check: 1 passed. Full-access parity across normal, channel, and teammate conversations: 8 passed.
• Inspected light/dark screenshots, model/effort selection, slider endpoints, and compact footer bounds. Browser checks use the existing mock daemon.
• Markdown links, public-repository safety scan, and whitespace checks passed.
• Secret scanning over the same remote refs as CI passed (1,034 commits). Eleven historical release-manifest source checksums were independently recomputed against their source files and excluded using exact commit/file/rule/line fingerprints; no rule or scan scope was changed.

Broader timeline-anchor remeasurement checks have an existing failure: both variants also fail at the same assertion in a separate, unmodified checkout of the current main revision. These test cases and timeline implementation remain unchanged; the failures are retained in the local validation record. No assertion or timeout was relaxed.

Review and compatibility

Parent and native adversarial source reviews found no actionable issue; the native pass reviewed test/fixture changes in summary mode. Documentation review confirmed the updated composer and full-access contracts. Outside-provider review was not completed.

Reasoning-summary workspace defaults and restored values still reach outgoing requests. Configuration remains workspace-scoped; no application-wide settings backend, web-search/output-detail options, API/schema, provider runtime, or native platform changes are introduced. The Alpha version is unchanged, with an Unreleased changelog entry. This PR does not deploy or restart the installed application.

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	🔄 Running since 2026-10-23T12:37:52.525733Z	59735aa	New commits
🔒 Security Review	⚠️ Failed 2026-10-23T12:41:11.723246Z We couldn’t complete this review. Comment @codex security review to try again.	59735aa	New commits

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.