feat(shell): open your profile from the account menu avatar

[matt2e]

Open your own community profile from the account menu avatar using the existing companion panel. Preserve keyboard navigation and focus return, and keep the avatar presentational in Personal space or when no profile panel is available.

Adds component and app wiring regression coverage and updates the shell design documentation.

Current validation — abb51df1

Integrated main 986f6881 (including #327) without changing the feature beyond the automatic merge. Independent source review found the four previous findings addressed; the integrated wiring was rechecked.

• Full Vitest: 447 files, 5,401 tests passed.
• Chromium + WebKit: 32 cases passed, no retries, via bin/pnpm test:browser account-profile.spec.mjs settings.spec.mjs profiles.spec.mjs --project chromium --project webkit --no-deps. The path patterns also include channel-settings and notification-settings.
• Required staged checks, TypeScript/related-unit checks, design checks and whitespace check passed. DCO trailer audit covers all outgoing feature commits.
• Browser accounting: 2 new account-profile cases, 0 removed (4 executions across the two engines). These cover real menu-close focus handoff and Escape routing for unavailable first-open and retained reopening. The existing Settings journey is updated, not replaced. Regression failure evidence for the earlier intermittent mounted test is in the resolved review thread; fail-then-pass for the new browser cases was not independently reproduced.

Pending: new-head hosted CI/DCO and human acceptance; browser tests are not native-app acceptance. No approval or merge was performed by the integrating agent.

Try: in a selected community, choose View your profile from the account-menu avatar, reopen with the panel already open, and press Escape. Focus should return to the header avatar. Repeat after a failed connection; opening/closing from New message should retain the recipient and unsent draft.

Profile-menu demo — 85eff698, Chromium, synthetic data. Open from a channel or New message; retain the recipient, draft and Channels tab across mouse/keyboard reopenings; close with Escape and continue editing. Capture assertions passed; native acceptance and human testing remain unverified.

profile-menu.mp4

[wesbillman]

Star Lord automated source review — via Wes's account

Reviewed head 85eff698db565f11199f3d913f2636f5ba409153 against pinned base df4b735c8bfa98018ee4e65de60b5fb8fd0e6f31.

Three actionable comments below: one first-open failure-state focus defect, one stale Settings keyboard journey, and one unit-fixture incompatibility demonstrated by the hosted merge build. The existing profile/session owner and retained companion approach fit the requested scope; no parallel data owner is needed. This is a non-blocking COMMENT review, not approval or merge authorization.

Validation and public-material limits

• Source-only: no PR code, tests, app, browser flow, or native build executed locally. Inspected callers, retained-panel lifecycle, initial/unavailable focus paths and existing profile retry handling.
• Hosted snapshot: run 36517276051 failed JavaScript and Chromium/WebKit journey shard 2. JavaScript recorded 419 passing files / 5,089 passing tests, with the new avatar assertion as the sole failure; both browser failures stop at the old Settings keyboard-focus expectation. This run tested merge ace8198389d7500cbf0fbd2f14fb4084b34a9ca5 (this head merged into cfbb61246425cbfd18c10c33ab96d63fedf17942), not the isolated reviewed head/base. I inspected the relevant merge-source differences rather than attributing its presence behavior to the branch source.
• Public repository: inspected the changed code/tests/docs, commit messages, and current PR description. No internal coordination URL, deployment identifier, or secret found in those inspected materials. The description now attaches an MP4 demo; it was downloadable but could not be visually inspected with the available tools, so its visual/privacy surface remains unverified. No image attachments were present in the description.
• The PR description still needs the repository-required browser-case accounting/justification and fail-then-pass evidence (one browser case added, none removed). Browser focus/keyboard routing is the stated browser-only boundary. Hosted repairs, native acceptance, and human confirmation remain outstanding; no buzz-review-completed attestation is established here.

[wesbillman]

Carl, an automated reviewer, commenting via Wes’s GitHub account.

No remaining code blockers found in this follow-up. Reviewed head 309d5228ac67e877400558410db4292a56324251 against base c6b47a5837fd8912dc84d98bb959714a23d70818. The earlier unavailable-profile focus, Settings keyboard sequence, presence fixture, and retained-reopen test-ordering findings are addressed; all four existing review threads are resolved. The final test change waits for actual menu focus rather than adding a delay or retry, consistent with Base UI’s deferred initial-focus implementation.

Hosted CI run 36538214625 is green: 420 Vitest files / 5,091 tests, Rust/tool integration, all Chromium/WebKit journey shards, measurements, and DCO/security checks. Logs confirm both new profile journeys and the existing Settings journey pass in both engines. That run checked merge 22b6d8f8da8c021de496bd8beec8c0246a40753f of the exact head/base above. I reviewed source and existing CI evidence; I did not rerun tests or exercise the native app locally. The clean head also passes git diff --check.

Ready for human approval on code quality; merge-readiness still needs human smoke-test confirmation. The PR description explicitly leaves human testing pending. Confirm avatar → own profile → repeat opening → Escape/focus return, including retaining a New message recipient/draft, then update the readiness checklist. The repair commit records browser-case accounting and fail-then-pass evidence, but the PR description has not been refreshed. These are readiness/documentation gaps, not additional code findings. Native acceptance remains unverified. This is a COMMENT review, not approval or merge authorization.

Public-material check: inspected changed source/tests/docs, commit text, PR description, and sampled frames from the attached synthetic-data demo; no disclosure issue found in those inspected materials. The demo is from an earlier head, not current-head runtime validation.