feat(tui): warn when creating a panel with no experts

[pedrofuentes]

PanelCreateScreen now shows a clear teacher ('No experts yet — create an expert first') instead of a silent empty member list when there are no experts; Phase 11 polish — feedback #2

[pedrofuentes]

Status: REJECTED

Sentinel Review Report

Ref: feature/tui-create-noexperts → main
Report ID: SENTINEL-1754-eaa70f1
Reviewed SHA: eaa70f1
Sentinel ruleset: v1
Reviewed at: 2026-06-26T10:25:36-07:00
Mode: standard
Review depth: Tier 2 (full)
Required action: FIX_AND_REINVOKE
Persisted report: #1754 (comment)

Phase 1 — TDD / Test Evidence

• Tests exist & meaningful: ❌ Partial. The warning render tests are meaningful and fail pre-implementation, but the focus-advance behavior is not tested (see 🔴 finding).
• Test-first history verified: ✅ git log --oneline origin/main..origin/feature/tui-create-noexperts shows 5e48e2b test(tui): ... before eaa70f1 feat(tui): ...; per-commit diffs show the test commit changed only packages/cli/tests/unit/tui/panel-create-screen.test.tsx, and the feat commit changed only packages/cli/src/tui/screens/PanelCreateScreen.tsx.
• Full suite green on SHA: ✅ Scoped final evidence at reviewed SHA: tests/unit/tui passed 95 files / 902 tests; focused panel-create-screen.test.tsx passed 12/12; lint and typecheck exited 0.
• Coverage: N/A (coverage threshold not requested/enforced for this scoped review).
• Test fails without implementation: ✅ At test commit 5e48e2b, panel-create-screen.test.tsx failed 2/12 on missing No experts yet assertions, proving the warning tests exercise the real screen.

Phase 1.5 — Fast-path Evaluation

🔴 count: 0 | LOC: 26 non-test insertions / 14 non-test deletions (≤150: Y) | Security paths: Y (terminal UI rendering surface) | New deps: N | Commit types qualify: N (feat)
→ Fast-path eligible: NO → Phase 2

Phase 2 — Execution Log

Commands run:

• git fetch origin && git log --oneline origin/main..origin/feature/tui-create-noexperts && git diff origin/main...origin/feature/tui-create-noexperts
• git show --patch 5e48e2ba1710708d625c023bdb2c04c6ee9f06ad -- packages/cli/tests/unit/tui/panel-create-screen.test.tsx
• git show --patch eaa70f18d880eca3e60376fb0da2964567d13133 -- packages/cli/src/tui/screens/PanelCreateScreen.tsx
• git worktree add --detach .worktrees/sentinel-1754-review eaa70f18d880eca3e60376fb0da2964567d13133
• npx --yes pnpm@10.30.1 install --frozen-lockfile (needed after initial run failed with missing kysely module in the review worktree)
• cd packages/cli && npx --yes pnpm@10.30.1 exec vitest run tests/unit/tui → ✅ 95 test files passed; 902 tests passed
• cd packages/cli && npx --yes pnpm@10.30.1 lint → ✅ eslint src tests, exit 0
• cd packages/cli && npx --yes pnpm@10.30.1 typecheck → ✅ tsc --noEmit, exit 0
• git checkout 5e48e2b && cd packages/cli && npx --yes pnpm@10.30.1 exec vitest run tests/unit/tui/panel-create-screen.test.tsx → ✅ expected RED: 1 file failed; 2 failed / 10 passed / 12 total
• git checkout eaa70f1 && cd packages/cli && npx --yes pnpm@10.30.1 exec vitest run tests/unit/tui/panel-create-screen.test.tsx → ✅ 1 file passed; 12 tests passed
• gh pr comment 1754 --body-file - → persisted report: feat(tui): warn when creating a panel with no experts #1754 (comment)

Dimension dispatch:

Dim	Tool Call	Agent ID / Ref	Status
Quick scan	task(agent_type="general-purpose", model="claude-haiku-4.5", name="sentinel-fast-scan")	N/A (sync task tool did not expose an agent id)	✅ 0 🔴; escalated to Tier 2 because fast-path criteria failed
A1	task(agent_type="general-purpose", name="sentinel-dim-a1")	N/A (sync task tool did not expose an agent id)	✅ No findings
A2	task(agent_type="general-purpose", name="sentinel-dim-a2")	N/A (sync task tool did not expose an agent id)	✅ No findings
B	task(agent_type="general-purpose", name="sentinel-dim-b")	N/A (sync task tool did not expose an agent id)	✅ No findings
C	task(agent_type="general-purpose", name="sentinel-dim-c")	N/A (sync task tool did not expose an agent id)	✅ No findings
D	task(agent_type="general-purpose", name="sentinel-dim-d")	N/A (sync task tool did not expose an agent id)	✅ No findings
E	N/A	N/A	N/A (no dependency surface changed)
F	task(agent_type="general-purpose", name="sentinel-dim-f")	N/A (sync task tool did not expose an agent id)	✅ Returned focus-control issue; orchestrator reclassified as correctness/TDD blocker

Findings

• 🔴 CRITICAL: 1
• 🟡 IMPORTANT: 0 new / 0 known
• 🟢 MINOR: 0

Details (ordered by severity)

1. 🔴 Empty-experts focus guard is incomplete and untested — packages/cli/src/tui/screens/PanelCreateScreen.tsx:88-90,130-133; packages/cli/tests/unit/tui/panel-create-screen.test.tsx:257-287
   • Evidence:

     if (key.tab) {
       setField((current) => (current === "name" ? "members" : "name"));
     }

     if (!hasNoExperts) {
       setField("members");
     }

     expect(lastFrame()).toContain("No experts yet");
     expect(lastFrame()).toContain("create an expert");

   • Impact: Trigger: a user creates a panel while experts.loadList() returns [] and presses Tab (the UI still advertises Tab focus). Mechanism: the global useInput Tab handler still sets field to "members" without checking hasNoExperts, while the new guard only covers TextInput.onSubmit. Consequence: focus can still advance to an empty/nonexistent Members control, violating the PR intent and leaving the new focus behavior without a regression test.
   • Remediation: Gate the Tab focus transition when hasNoExperts is true (or keep focus on name whenever no member list is rendered) and add a no-experts keyboard test that writes \t and/or \r and asserts focus does not move to Members.

Follow-ups & Actions

• REJECTED → FIX_AND_REINVOKE: fix 🔴 blockers only, re-commit, re-invoke Sentinel. Do not merge.

Decision rationale

• TDD ordering and commit purity are compliant.
• The warning copy is a trusted literal; terminal sanitization is N/A for that new string, and existing dynamic terminal strings in this screen use toSingleLineDisplay.
• The submit guard requiring at least one selected expert remains unchanged.
• Final scoped tests, lint, and typecheck pass at eaa70f18d880eca3e60376fb0da2964567d13133.
• The PR’s stated focus behavior is incomplete because Tab still reaches the empty Members field and no test covers that path.

Verdict: REJECTED (reviewed SHA: eaa70f1)

[pedrofuentes]

Status: APPROVED

Sentinel Review Report

Ref: feature/tui-create-noexperts → main
Report ID: SENTINEL-1754-9a49a15
Reviewed SHA: 9a49a15
Sentinel ruleset: v1
Reviewed at: 2026-06-26T10:43:42-07:00
Mode: standard
Review depth: Tier 2 (full)
Required action: MERGE
Persisted report: #1754 (comment)

Phase 1 — TDD / Test Evidence

• Tests exist & meaningful: ✅ New regression test exercises the Tab focus path by typing after Tab and asserting the name field still contains the new character: expect(lastFrame()).toContain("abcd"); (packages/cli/tests/unit/tui/panel-create-screen.test.tsx:291-305).
• Test-first history verified: ✅ Chronological history shows e1fbf75 test(tui): assert Tab keeps focus on name when no experts before 9a49a15 fix(tui): keep focus on name when creating a panel with no experts; per-commit file lists confirm e1fbf75 changed only packages/cli/tests/unit/tui/panel-create-screen.test.tsx and 9a49a15 changed only packages/cli/src/tui/screens/PanelCreateScreen.tsx.
• RED→GREEN proof: ✅ At e1fbf75, vitest run tests/unit/tui/panel-create-screen.test.tsx failed with AssertionError: expected 'Name: abc\n\u001b[7mMembers:\u001b[27…' to contain 'abcd' and 1 failed | 12 passed (13); at 9a49a15, the same file passed 13 passed (13).
• Relevant suite green on SHA: ✅ tests/unit/tui passed 95 passed (95) test files and 903 passed (903) tests at 9a49a15c62afe58019e7145aa71c72aa69c43f6f; npx --yes pnpm@10.30.1 lint and npx --yes pnpm@10.30.1 typecheck both exited 0.
• Coverage: N/A (not enforced for this scoped re-review; no coverage threshold command requested).

Phase 1.5 — Fast-path Evaluation

🔴 count: 0 | LOC: 4 non-test lines changed (≤150: Y) | Security paths: Y (terminal UI surface touched) | New deps: N | Commit types qualify: Y
→ Fast-path eligible: NO → Phase 2

Phase 2 — Execution Log

Dim	Tool Call	Agent ID / Ref	Status
Quick scan	task(agent_type="explore", name="sentinel-quick-scan", model="claude-haiku-4.5")	N/A (tool returned sync result, no ID)	✅ 0 🔴
A1	task(agent_type="general-purpose", name="sentinel-dim-a1", model="claude-sonnet-4.6")	N/A (tool returned sync result, no ID)	✅ No findings
A2	task(agent_type="general-purpose", name="sentinel-dim-a2", model="claude-sonnet-4.6")	N/A (tool returned sync result, no ID)	✅ No findings
B	task(agent_type="general-purpose", name="sentinel-dim-b", model="claude-sonnet-4.6")	N/A (tool returned sync result, no ID)	✅ No material scoped findings
C	task(agent_type="general-purpose", name="sentinel-dim-c", model="claude-sonnet-4.6")	N/A (tool returned sync result, no ID)	✅ No findings
D	task(agent_type="general-purpose", name="sentinel-dim-d", model="claude-sonnet-4.6")	N/A (tool returned sync result, no ID)	✅ No material scoped findings
E	N/A	N/A	✅ Auto-skip: no dependency/supply-chain surface changed
F	task(agent_type="general-purpose", name="sentinel-dim-f", model="claude-haiku-4.5")	N/A (tool returned sync result, no ID)	✅ No material findings
Persist	gh pr comment 1754 --body-file .sentinel/reports/SENTINEL-1754-9a49a15.md	#1754 (comment)	✅ Full report posted to PR

Commands run:

• git fetch origin && git --no-pager status --short --branch && git --no-pager log --oneline origin/main..origin/feature/tui-create-noexperts && git --no-pager diff --name-only eaa70f1...origin/feature/tui-create-noexperts && git --no-pager diff --stat eaa70f1...origin/feature/tui-create-noexperts
• git --no-pager diff eaa70f1...origin/feature/tui-create-noexperts -- packages/cli/src/tui/screens/PanelCreateScreen.tsx packages/cli/tests/unit/tui/panel-create-screen.test.tsx
• git --no-pager show --stat --patch --find-renames e1fbf75 and git --no-pager show --stat --patch --find-renames 9a49a15
• git checkout e1fbf75 && cd packages/cli && npx --yes pnpm@10.30.1 exec vitest run tests/unit/tui/panel-create-screen.test.tsx → expected FAIL (1 failed | 12 passed (13))
• git checkout 9a49a15 && cd packages/cli && npx --yes pnpm@10.30.1 exec vitest run tests/unit/tui/panel-create-screen.test.tsx → PASS (13 passed (13))
• git rev-parse HEAD && cd packages/cli && npx --yes pnpm@10.30.1 exec vitest run tests/unit/tui && npx --yes pnpm@10.30.1 lint && npx --yes pnpm@10.30.1 typecheck → PASS (95 passed (95), 903 passed (903))
• git --no-pager grep -n -E 'setField|"members"|hasNoExperts|noExperts|warning|No experts' origin/feature/tui-create-noexperts -- packages/cli/src/tui/screens/PanelCreateScreen.tsx

Findings

• 🔴 CRITICAL: 0
• 🟡 IMPORTANT: 0 new / 0 known
• 🟢 MINOR: 0

Details (ordered by severity)

1. [RESOLVED 🔴] Previous no-experts Tab focus leak — packages/cli/src/tui/screens/PanelCreateScreen.tsx:88-92
   • Evidence: const noExperts = expertState.status === "loaded" && expertState.data.length === 0; / if (!noExperts) { / setField((current) => (current === "name" ? "members" : "name"));
   • Impact: The prior rejected SHA allowed Tab to focus the absent members control when the expert list was loaded-empty; RED proof showed the new test failed before the fix.
   • Remediation: Resolved by guarding the Tab focus sink with the same loaded-empty source used by the no-experts render path.
2. [RESOLVED 🔴] Exhaustiveness of focus-to-members sinks — packages/cli/src/tui/screens/PanelCreateScreen.tsx:133-137
   • Evidence: onSubmit={() => { / if (!hasNoExperts) { / setField("members");
   • Impact: The Enter/onSubmit focus sink remains guarded, and grep found no other setField sink to "members" beyond the guarded Tab and onSubmit paths.
   • Remediation: Verified no remaining unguarded loaded-empty focus path to members.
3. [Verified] With-experts Tab behavior and warning/sanitization — packages/cli/src/tui/screens/PanelCreateScreen.tsx:143-147
   • Evidence: props.theme.warn( / "No experts yet — press Esc, then create an expert (Experts → n) before building a panel.",
   • Impact: Warning copy is a trusted literal; no new unsanitized dynamic terminal string was introduced. Existing with-experts test still passes through Tab/member selection, and the final TUI suite passed.
   • Remediation: None required.

Follow-ups & Actions

• APPROVED → MERGE: no new 🟡/🟢 issues to file from this scoped re-review.
• ⚠️ Do NOT fix 🟡/🟢 findings in this PR — none were accepted as material scoped findings.

Decision rationale

• The new TDD pair is correctly ordered and pure: test-only commit e1fbf75 precedes implementation-only commit 9a49a15.
• The new test genuinely fails without the fix and passes at the reviewed SHA, proving regression coverage for the rejected focus path.
• Both loaded-empty focus-advance sinks to members are guarded (Tab via noExperts; Enter/onSubmit via hasNoExperts), and grep found no additional setField("members") path.
• Relevant TUI tests, lint, and typecheck all pass at the reviewed SHA.
• Phase 2 found no material in-scope security, resilience, performance, testing, dependency, or documentation blockers.

[pedrofuentes]

Sentinel re-review SENTINEL-1754-9a49a15 → APPROVED (0 🔴/🟡/🟢). Tab + onSubmit focus sinks both guarded against the no-experts case; RED→GREEN proven. Merging.