Skip to content

app: persistent agent strip in Diff tab + board-card severity escalation#29

Merged
alejandro-nexcade merged 3 commits into
mainfrom
alejandro/diff-strip-card-chips
Jul 3, 2026
Merged

app: persistent agent strip in Diff tab + board-card severity escalation#29
alejandro-nexcade merged 3 commits into
mainfrom
alejandro/diff-strip-card-chips

Conversation

@alejandro-nexcade

Copy link
Copy Markdown
Collaborator

What

PR 5/7 of the overnight UX program (mockups A2 + A3 — cockpit-docs/OVERNIGHT_UX_PLAN.md). Stacked on #28.

  • AgentStrip: a persistent, never-collapsible one-liner under the Diff-tab header — ✦ AGENT'S READ eyebrow, plain-text intent preview, non-zero severity chips (dot+label), and BRIEFING ↗ switching to the Briefing tab. Renders nothing when the review has no agent data. Props-only, reusable by PR 7's preview rail.
  • Band surgery: IntentPanel and EvidenceStrip bands removed from DiffView (their content lives in the Briefing hero + InstrumentRow since app: Briefing tab — intent, instruments, findings, and visual recap as the review's front page #28); both components and their tests deleted as verified-dead code. ConversationBand, FindingsPanel (auto-expand-on-critical intact), AddressedRequests, and the Stack strip stay.
  • Board cards: CRIT/WARN severity chips in both densities (shared chip anatomy) + reason escalation "AGENT FLAGGED N CRITICAL" in the danger tone, slotted stale > CI-failing > critical-findings > large-diff > sensitive-paths (documented in card-signal.ts; warnings never escalate).

Verified

tsc clean · vitest 319 (29 files; new AgentStrip/card-signal/finding-severity suites, ReviewCard tests extended additively) · vite build ok · grep-verified zero dangling references to deleted components. Reviewer pass: SHIP; both NITs addressed (stale comments reworded; brand-colored BRIEFING link kept — it matches the approved mockup's agent-voice treatment).

🤖 Generated with Claude Code

@LifeLex LifeLex force-pushed the alejandro/briefing-tab branch from d6acce6 to 4b79f0d Compare July 3, 2026 08:37
@alejandro-nexcade alejandro-nexcade changed the base branch from alejandro/briefing-tab to main July 3, 2026 08:40
LifeLex and others added 3 commits July 3, 2026 09:40
…ands to Briefing

Mockup A2. Add an always-visible, never-collapsible AgentStrip directly under
the Diff-tab header: the one-line "agent's read" (markdown stripped, never
rendered), live severity chips (n CRIT/WARN/INFO, non-zero only), and a link to
the full Briefing. Renders nothing when the review has neither intent nor
findings. Props-driven with no store reads so PR 7's board preview rail can
reuse it.

Remove the IntentPanel and EvidenceStrip bands from DiffView — their content now
lives on the Briefing tab (hero + instrument row). Both components became fully
unused (grep confirms no other importers; EvidenceStrip's classifyAgentRan /
isVerificationCommand had no external users), so the component and test files are
deleted. Deleting tests of deleted components is dead-code hygiene, not
test-weakening. FindingsPanel stays and remains reachable on its own; its
auto-expand-on-critical is unchanged (findingsAutoExpand). Now-dead DiffView
state/effects (intent disclosure, evidence fetch, raw ci-checks, findings-panel
scroll ref) are cleaned up; evidence fetching is dropped here since only the
removed strip used it (the new strip's severity chips read review.review_findings).

Add findingCounts() to finding-severity.ts — a pure, dismissal-agnostic
per-severity counter shared by the strip (and the board cards).

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
…ation

Mockup A3. Add advisory-finding severity chips to the card telemetry/risk row in
both densities (n CRIT danger-toned, n WARN warning-toned; INFO omitted on
cards), matching the existing chip anatomy (h-1.5 dot + mono tabular-nums label).
Counts come from review.review_findings via the shared findingCounts helper — no
new IPC; the board's existing refresh path repaints cards.

Escalate the reason line to danger tone (red dot + red uppercase text) —
"Needs your review · Agent flagged N critical" — when a review has ≥1 Critical
finding. Slotted directly below CI-failing and above the softer size /
sensitive-path notes (documented in card-signal.ts): a broken build stays the
headline (chip only for criticals then), stale still preempts everything, and
warning-only findings never escalate (chips only). Adds a "danger" SignalTone.

Existing card-signal / ReviewCard expectations are unchanged except where the new
spec applies; no assertions were loosened.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
@LifeLex LifeLex force-pushed the alejandro/diff-strip-card-chips branch from 1068e2f to f08887c Compare July 3, 2026 08:40
@alejandro-nexcade alejandro-nexcade merged commit 6e649d0 into main Jul 3, 2026
2 checks passed
@LifeLex LifeLex deleted the alejandro/diff-strip-card-chips branch July 3, 2026 08:48
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants