PR 19859, round two: the API Key fix

Your PR again, now at 5227ba5 ("Show a cloud credential's public key as text"). Reviewed twice, by Cody's round-two process (my-pr-review-new-commits, ai-skills c04c11a) and by ours, each in a fresh agent, then merged and re-checked. Nothing was posted to GitHub.

Preview: rancheri19854r, rebuilt from 5227ba5, backend the mf Rancher (v2.16 head). Baseline: dashboard-master rebuilt again at 55438bb (still the upstream tip), same backend. Add a DigitalOcean credential, then View > Group By > API Key.

Where it stands: ready, nothing to fix in the code

Both passes agree that round one's regression is fixed where the value is made, and that the new commit adds nothing that needs changing. The first commit's fix still holds. Two optional nits (1, 2) and one FYI (3).

Left to merge, none of it code: the two checklist boxes Richard asked about, the code-freeze exception, an a11y-test rerun once the runner is healthy, and the 2.17.0 issue for long values.

Round one, now

Round oneWhat changedChecked on 5227ba5Now
API Key header showed &hellip; (both passes)publicData returns plain text with a real …; the cell shows it with {{ }}. New tests fail if either change is undoneBoth passes, each with its own throwaway credentials: the header equals the cell and master for a plain DigitalOcean key, and for keys holding & < > " ' in prefix and full mode. No tags rendered. Pass B's header crops are byte-identical to master'sFixed at the cause
No 2.17.0 issue for long valuesNothingSearched issues created since 11-01: noneStill open
CI: Description"Reviewer assigned" now tickedAccessibility and Global Roles still unticked; Richard says both can be tickedStill open
CI: Branch MilestoneNothingMilestone matches; fails on master's code freezeNeeds the freeze exception
CI: a11y-testNothing503 at "Run Rancher" on both attempts, and on three other PRs this morningInfrastructure
At ce68347: the API Key group header reads i19854pb&hellip; while the cell reads i19854pb….
Round one, at ce68347: the header shows the entity.
At 5227ba5: a key a&<b>"cd… reads the same in the header and the cell.
Now, at 5227ba5: a key with & < > " reads the same in the header and the cell.

How the two passes compare

#PointPass A (Cody)Pass B (ours)Lands as
1The full-mode test title reads "shows the full of the key"-N1, codeOptional
2The filter, search and export for API Key change too, and the description doesn't say soSeen as an improvement, not filedN2, liveOptional
3A key holding " now hits issue 19853 when picked from the filter suggestions-Outside the new commitFYI
Pass A (Cody's process, round two)Pass B (our process, round two)
VerdictLooks ready to approve: 1 of 1 Addressed, no new commentsReady: no blockers, no suggestions, 2 optional nits
Checked by bothThe round-one steps on the rebuilt preview against master with throwaway DigitalOcean and AWS credentials, plain and with special characters. The first commit's ConfigMaps fix still holding. Every other reader of publicData (none in shell or pkg; none found in the extension repos). The CI state.
Only one pass didDropped with reasons: sort order follows what's shown, no saved views to break yet, the formatter's nullable prop is master's.Four mutants (escaping back in each mode, v-clean-html back in the cell), all killed. Picking filter suggestions, the CSV export, sort. Lint and type-check. A coverage grid.
Outputreview.md, the state comment, evidencereview.md, coverage.md, checked.md, draft.md, screenshots
CostA fresh agent: 13 minutes, 62 tool calls, about 151k tokensA fresh agent: 22 minutes, 101 tool calls, about 195k tokens

New points

1

The full-mode test title reads "shows the full of the key"

Found by Pass B (ours)
Optional

The it.each title puts the mode name in %s, so the full-mode case prints "shows the full of the key as plain text". The prefix and suffix cases read fine, and the test itself is right (it kills the mutant that escapes a full key again). Only the wording in the jest output.

2

The filter, search and export for API Key change too, and the description doesn't say so

Found by Pass A (Cody)Found by Pass B (ours)
Optional

The same value feeds the apikey: filter suggestions, free-text search and Export All View Contents. On master they got entity text: typing hellip matched every DigitalOcean credential, typing a&b<i> matched nothing, and the suggestion and CSV read i19854pb&hellip;. On the PR all three use the key as typed. That's better, but the description only mentions the header and the cell, so QA won't look at the rest.

3

A key holding " now hits issue 19853

Found by Pass B (ours)
FYI

Picking a filter suggestion for a key that contains a double quote shows no rows on the PR. Master only avoided it because it encoded the quote as &quot;. This is izaac's open issue 19853 ("Filter suggestion with double quotes returns no rows", v2.17.0), not something this commit gets wrong, and real cloud keys don't contain quotes. A key with & < > and no quote filters correctly on both. Noted so nobody reads it as a regression from this PR.

Also seen

What I'd do

The code is done. If you want the nits: reword the test title (1) and add a line about the filter, search and export to the description (2). Then the non-code items: tick Accessibility and Global Roles and answer Richard, file the 2.17.0 issue, and rerun a11y-test.