umbrella: frontend coverage debt — 11 files have shallow tests, blocked by missing Playwright suites

[Kpa-clawbot]

Frontend coverage debt — 11 HIGH-priority files have shallow/zero test coverage

Aggregate frontend coverage today: 40.02% statements (latest .badges/frontend-coverage.json).

Audit report: notes/frontend-coverage-audit-2026-05-19.md (workspace-local; copy-pasted summary below).

Goal

70%+ statements per file for the 11 🔴 HIGH-priority files (~9.4k LOC). Estimated lift on the aggregate badge: 40% → ~55–60%.

Approach

Parallel fixer subagents, each owning one batch (3–5 files of related surface). Each batch lands as a single PR adding Playwright E2E suites (plus unit tests where the surface is pure logic). One umbrella tracks five sub-tasks.

Counts (from audit)

• 🔴 HIGH: 11 files
• 🟡 MED: 24 files
• 🟢 LOW: 6 files

Per-batch sub-tasks

• Batch 1 — audio-batch: audio.js, audio-v1-constellation.js, audio-lab.js. E2E: navigate /#/audio-lab, toggle modes, assert AudioContext + visualizer DOM; for audio.js inject mock packets and assert mute toggle.
• Batch 2 — mobile-chrome-batch: bottom-nav.js, nav-drawer.js, touch-gestures.js, drag-manager.js. E2E at mobile viewport (375x812): tap nav, swipe drawer, drag customizer cards; assert route changes, drawer aria, reorder persistence.
• Batch 3 — observer-batch: observer-detail.js, traces.js, node-analytics.js. E2E: visit each page-module's route with a seeded id, exercise tabs and panels.
• Batch 4 — customizer-batch: customize-v2.js (1774 LOC — likely split into 2–3 E2E suites: presets, column editor, persistence).
• Batch 5 — util-batch: table-sort.js. Multi-page E2E clicking <th> on every table-bearing route + comparator unit tests.

Top 5 most-impactful HIGH files

1. customize-v2.js (1774 LOC) — single biggest under-tested surface
2. touch-gestures.js (455 LOC) — gesture math, chrome-loaded
3. nav-drawer.js (366 LOC) — drawer a11y/focus-trap
4. observer-detail.js (355 LOC) — page module, zero tests
5. bottom-nav.js (323 LOC) — primary mobile nav, zero direct tests

Notes for the fixer chain

1. Run real coverage first. Spin up the Go server on :13581, run node scripts/collect-frontend-coverage.js, overlay real per-file % onto the HIGH list — it may demote or promote a file.
2. customize.js (1510 LOC, NOT in index.html script tags) is dead — confirm and delete in a separate PR rather than testing it.
3. Chrome-loaded files require driving them from multiple pages; page-module files only need their one route. Don't conflate.
4. PRs should land via the standard pipeline (fix-issue → CI → pr-polish), one batch per subagent, in parallel.

Reference

• Audit report (workspace): notes/frontend-coverage-audit-2026-05-19.md
• Coverage collector: scripts/collect-frontend-coverage.js
• Current badge: .badges/frontend-coverage.json

[Kpa-clawbot]

Refined priorities — live coverage measurement (2026-05-20)

The static-mention audit in the prior comment was based on grep hits across test-*.js. Re-ran with real per-file statement coverage from an instrumented public-instrumented/ against the e2e fixture (132s, 7 parallel browser groups via scripts/collect-frontend-coverage.js). Aggregate from collector alone: 34.95% stmt (the badge's 40% includes the full 45-suite CI run).

Static audit got the BIG things wrong

7 mis-prioritizations large enough to change batch composition:

file	LOC	static said	live shows	corrected
audio-lab.js	562	🟢 LOW (claimed e2e-covered)	4.18%	🔴 HIGH
channel-qr.js	280	🟡 MED	2.29%	🔴 HIGH
channel-decrypt.js	439	🟡 MED (8 unit tests!)	8.54%	🔴 HIGH
channel-color-picker.js	284	🟡 MED	6.62%	🔴 HIGH
path-inspector.js	205	🟡 MED	4.72%	🔴 HIGH
channels.js	1950	🟡 MED	19.93% (1036 uncovered stmts)	🔴 HIGH
home.js	564	🟡 MED	20.91%	🔴 HIGH
bottom-nav.js	323	🔴 HIGH	61.44%	🟢 LOW (drop)
observer-detail.js	355	🔴 HIGH	70.08%	🟢 LOW (drop, at target)
nav-drawer.js	366	🔴 HIGH	42.91%	🟡 MED (downgrade)
table-sort.js	224	🔴 HIGH ("ZERO tests")	55.41%	🟡 MED (downgrade)
drag-manager.js	216	🔴 HIGH	25.13%	🟡 MED
audio.js	214	🔴 HIGH	33.05%	🟡 MED

Pattern: "test-mention count" massively overweights regression unit tests that touch ~5% of a file. And "ZERO mentions" was wrong for several files exercised transitively by page-level e2e tests (bottom-nav, observer-detail, table-sort).

Confirmed HIGH (no change)

audio-v1-constellation (8.3%), traces (8.5%), touch-gestures (14%), customize-v2 (16.1%, biggest absolute deficit: 1488 uncovered stmts), node-analytics (16.3%).

Refined batch breakdown

batch	composition	change vs static audit
B1 audio	audio-lab + audio-v1-constellation + audio.js	audio-lab promoted from LOW to focus of the batch
B2 channel-decode chrome (NEW)	channel-decrypt + channel-qr + channel-color-picker	brand new batch — all three under 10% live
B3 channels page	channels.js (+ light touch on observers.js)	channels promoted MED→HIGH
B4 customize-v2	customize-v2 + drag-manager	nav-drawer/bottom-nav dropped (already 43-61%)
B5 page modules	traces + node-analytics + path-inspector + home	observer-detail dropped (70% already); home + path-inspector promoted
B6 touch-gestures	touch-gestures only	table-sort, nav-drawer, bottom-nav all dropped

Net: the static audit's "Mobile chrome / gestures" batch (B2) is essentially gutted. The real fat is in channel-decode chrome (a batch that didn't exist before) and the customize-v2 + channels + analytics megafiles.

Top 3 misprioritizations (impact-ranked)

1. audio-lab.js (562 LOC) marked LOW — actually 4.2%, the single biggest static-audit miss.
2. channels.js (1950 LOC) marked MED — 1036 uncovered stmts; should be a HIGH batch on its own.
3. bottom-nav.js (323 LOC) marked HIGH — already at 61%, would have been wasted subagent effort.

Full refined table (all 39 instrumented files) is in notes/frontend-coverage-audit-2026-05-19.md § "Refined coverage — live measurement 2026-05-20".