feat(profiles): show agent metadata and copyable nip05

[kalvinnchau]

Summary

• Show public agent type and capabilities using verified session metadata and managed-owner precedence.
• Add copyable NIP-05, agent type and capabilities rows with shared toast feedback, preserving the unverified identifier label and existing public-key copy behavior.
• Keep metadata withheld until ownership settles and recover through the existing profile retry control.

Verification

• Node integration: 133 passed; Vitest: 3,729 passed; Rust: 25 passed.
• TypeScript, formatting, design checks and pre-push hooks passed.
• Profile navigation, recovery, appearance and clipboard-boundary journeys passed in Chromium and WebKit.
• Full browser comparison: candidate 664/668 passed; clean main 666/668 passed. Kalvin authorized ready-for-review delivery with the aggregate results disclosed in the originating review thread.
• Extended the existing browser profile journey for keyboard copy affordances and host-toast wiring; metadata matrices remain in colocated tests.

[wesbillman]

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

Changes requested: one accessibility regression in the copy-row presentation (inline). Resolve it with coverage of the announced values and NIP-05 qualifier. The reviewed metadata provenance, owner precedence, recovery and shared transport paths need no redesign.

Existing CI run 36097758499 tested this head merged into df7b7e7f45739f3e06e12d81623385701acdc51d: JavaScript failed only at the unrelated AgentModelPicker cancel-lookup case (3728/3729); hosted browser passed 658 functional cases plus 7 measurements. That required CI gate remains separate from this finding. No rerun requested.

Source-only review on isolated Blox; no PR code executed. Ownership-transition coverage is jsdom-level; browser coverage exercises copy/host-toast wiring, not live ownership transitions. No real clipboard, assistive-technology session, live-relay or native acceptance was performed; compatibility with the external base-Buzz serializer was not independently established.

[wesbillman]

[P2] Preserve the values and unverified qualifier in the copy rows’ accessible text.

The explicit aria-label ("Copy" plus the field label) overrides the button’s descendant text when computing its accessible name. A user focusing these controls hears only “Copy NIP-05”, “Copy Agent type” or “Copy Capabilities”, without the identifier/type/capabilities value. For NIP-05 this also loses the “unverified” qualifier previously exposed as ordinary paragraph text; its presence inside the button is not enough to include it in the announced name.

Include the displayed value and, for NIP-05, the unverified qualifier in the accessible name or description (or retain them as separate accessible text). Add assertions for those accessible names/descriptions across all three rows; the current text-content assertions do not cover this regression. No metadata/lifecycle redesign is needed.

[wesbillman]

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

Clear on focused re-review; no remaining code blockers found. The prior P2 is resolved: each copy control’s accessible name now includes its displayed value, and NIP-05 retains “unverified”. Three exact computed-name assertions cover the regression. Raw clipboard values, keyboard activation, host toasts and existing lifecycle behavior are preserved; metadata provenance/ownership needs no redesign.

Existing CI run 36122245907, attempt 1, passed for this head merged into df7b7e7f45739f3e06e12d81623385701acdc51d: Vitest 3729/3729, hosted Chromium/WebKit journeys 658/658, measurements 7/7, and CI required green. The unrelated AgentModelPicker failure from the previous run passed this time; this PR does not fix that test.

Source-only review on isolated Blox; no PR code executed by reviewers. Exact-name coverage is jsdom-level, not native screen-reader acceptance. Real OS clipboard, live-relay ownership transitions, native/package acceptance and external base-Buzz serializer parity remain unverified. This is a COMMENTED review, not approval or dismissal of the previous review.