fix(desktop): restore the authored description on persona cards

[mfethe1]

Problem

Editing a built-in agent's description saves the value but never displays it. Upstream's regression test for block#1928 — agents.spec.ts:297 "built-in persona edits persist" — fails on our fork:

Locator: getByTestId('persona-agent-row-builtin:fizz')
Expected substring: "Helps teams ship reliably."
Received string:    "MFStartMy FizzDefault model (claude-opus-4-5)"

The name (My Fizz) persists; the description is simply absent from the card face.

Root cause

A sync merge left the call site in UnifiedAgentsSection.tsx passing card.personaLabel as the subtitle. That label is non-null only for a split card (disambiguating rivals); for a canonical card it is null (unifiedAgentGroups.ts:72), so runtimeSubtitle collapsed to empty.

Upstream computes that second line as effectiveAgentDescription(persona) ?? modelLabel. The helper still exists in our tree and is used elsewhere — only this call site lost it.

This is the fourth upstream guarantee broken by fork-local drift and surfaced by a byte-identical upstream spec.

Fix

Fall back to the authored description when no split label is supplied. Split cards keep their rival label; the Hermes · <profile> suffix is unchanged.

Evidence

Baseline established by running the same test on origin/product/main without this diff:

• agents.spec.ts:297 — FAILS on base, passes with this change
• agents.spec.ts full file (integration) — 37/37 pass
• desktop unit suite — 6583/6583 pass
• pnpm run check exit 0 · pnpm run typecheck exit 0

Note

Confirms PR #34 did not cause this integration failure — it reproduces identically on the base.

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on base/target branches other than the default branch.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Organization UI

Review profile: ASSERTIVE

Plan: Advanced

Run ID: 87d71d8f-11f3-46cb-a1a0-24870640fb44

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[mfethe1]

messaging.spec.ts:3105 is pre-existing on base — not caused by this PR

Differential run on this machine, fresh npm run build:e2e per ref, --project=smoke --retries=0, 3 runs each:

ref	run 1	run 2	run 3
origin/product/main (base)	1 failed	1 failed	1 failed
this branch (#36)	1 failed	1 failed	1 failed

Identical. The failure exists on the base branch without this diff, so it is not a regression from #36.

Correcting my earlier analysis

I previously called this "CI shard-order pollution" and said it passed in isolation. That was wrong, and the better test disproves it:

• I found a plausible polluter (message-feedback-snapshots.spec.ts run before :3105 reproduced Received: 58), but the same pair then passed 2/2 — so that was coincidence, not cause.
• :3105 also fails completely alone after a fresh build, with no other spec in the run.

There is no polluting test. The menuGap assertion (expect(menuGap).toBeLessThanOrEqual(headerPaddingInlineEnd + menuBox.width), 58 > 52) is environment-sensitive to layout/font metrics, and it fails that way on base too. The earlier "passes isolated" observations were environment-dependent, not code-dependent.

No layout code should be changed for this, and it should not gate this PR.