Skip to content

docs(status): the frontend audit is closed out, and what is left needs a person - #84

Merged
khafifithebork merged 1 commit into
masterfrom
docs/status-frontend-complete
Sep 28, 2026
Merged

khafifithebork merged 1 commit into
masterfrom
docs/status-frontend-complete

Conversation

@khafifithebork

Copy link
Copy Markdown
Owner

#83 merged, so every defect the audit found that was open work is fixed. The section's counts and PR table were a commit behind — but the more useful gap was that it never said plainly that there's no frontend work left an agent can do alone.

Docs only, one file.

What remains, and what each thing needs

Remaining Needs
The four unconfirmed claims A person with a browser. Minutes each
A2 — reproduce the 2.4.11 focus fault The same. The remedy can't be scoped until it's reproduced (ADR-006)
A5 — per-route first-load JS A measurement, before any performance budget can be proposed honestly
<track>'s srclang / label The API to model caption language — not knowable from the frontend
B2 — architecture.md §8.2 An owner decision: approve MSW / axe-core / Playwright under §5, or record that line superseded

Why the unconfirmed list has four rows and not forty

Worth stating, because without it the table reads either as alarming or as something to grind down with more tests — and it's neither.

Most of this work is logic a unit test exercises fully. D3's pagination has never been rendered in a browser either, and that's fine, because jsdom reaches every branch of it.

The four are different in kind: the property itself is invisible to the test environment. jsdom paints nothing, the agent's browser freezes its own clock, and a scrollbar has no API. No amount of further testing closes them — which is exactly what makes them a handover rather than a backlog.

Counts

Seven follow-up changes, 416 tests → 455 across 40 files, no dependency added, no service or cost changed.

Every ADR id and doc path resolves; no stale counts remain.

🤖 Generated with Claude Code

…s a person

#83 merged, so every defect the audit found that was open work is fixed. The
section's counts and PR table were a commit behind; more usefully, it did not say
plainly that **there is no frontend work left an agent can do alone.**

That is now a table, with what each remaining item needs: a browser and a person
for the four unconfirmed claims and for reproducing the 2.4.11 fault; a measurement
before any performance budget can be proposed honestly; backend modelling for
`<track>`'s language attributes; and an owner decision on architecture.md §8.2's
three uninstalled test dependencies.

**Also records why the unconfirmed list has four rows and not forty.** Most of this
work is logic a unit test exercises fully — D3's pagination has never been rendered
in a browser either, and that is fine, because jsdom reaches every branch of it.
The four are different in kind: **the property itself is invisible to the test
environment.** jsdom paints nothing, the agent's browser freezes its own clock, and
a scrollbar has no API. No amount of further testing closes them, which is what
makes them a handover rather than a backlog — and without that distinction the list
reads either as alarming or as something to grind down with more tests, and it is
neither.

Seven follow-up changes now, 416 tests to 455, no dependency added and no cost
changed.

Docs only. Every ADR id and doc path resolves; no stale counts remain.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
@khafifithebork
khafifithebork merged commit 7accb1b into master Sep 28, 2026
8 checks passed
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.

1 participant