Repository navigation
Agent chat tests: a prompt after tool rows stays centred in its gap - #31360
Conversation
Leo, 2026-11-01: "this is too close to tool calls, center it". Adds the gallery entry agent-pane.prompt-after-tool-rows and a Chromium test that the steer's bubble has as much room over it as under it. Red on base. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?id=aid-92fa78a91e29868c5eaa
Leo, 2026-11-01: "this is too close to tool calls, center it". A tool or one-line row right over the person's own prompt now ends at --cv-prompt-gap, the room the prompt has under its bubble (reply room + the next step's 2px top = 38px), instead of the 4px run gap. model.ts statusEdge marks it before-prompt and the estimate follows. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?id=aid-92fa78a91e29868c5eaa
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?id=aid-92fa78a91e29868c5eaa
|
cmux status (one comment, edited in place; each line is updated by the job that owns it) Review App of this head (app.zip, kept 1 day)
Gallery before/after for
States
Pick other states with a red: cmux-next checks (god files, concurrency, crash safety, l10n); no finished base run on 412578b to compare; do not merge until green; see lane-rules Shipped in:
|
|
Owned by an agent session · session |
|
Found 8 test failures on Blacksmith runners: Failures
|
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?id=aid-92fa78a91e29868c5eaa
Bugbot needs on-demand usage enabledBugbot uses usage-based billing for this team and requires on-demand usage to be enabled. A team admin can enable on-demand usage in the Cursor dashboard. |
#27580 moved the transcript onto stack gaps: a prompt takes the 32px turn gap on both sides, which centres it after tool rows. This branch's --cv-prompt-gap edge is dropped for base's; the gallery scene and the browser test stay as the guard, now asserting 32px both sides. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?id=aid-92fa78a91e29868c5eaa
Bugbot needs on-demand usage enabledBugbot uses usage-based billing for this team and requires on-demand usage to be enabled. A team admin can enable on-demand usage in the Cursor dashboard. |
|
Review (subagent opus-5.5 medium): SHIP at 4994a93 No blocking findings. The test measures what it says (nearest
|
![Fix with [code]smith](https://pr-comments-assets.blacksmith.sh/codesmith/fix-with-codesmith-light-3x.png)
Leo, 2026-11-01, on a steer whose bubble nearly touched the "Sent to …" tool line over it: "this is too close to tool calls, center it".
What this does
While this PR was open, #27580 landed. It moved the transcript onto stack gaps (grid.md §5.5): a prompt now takes the 32px turn gap on both sides, which already centres it after tool rows. This PR's original fix (a
--cv-prompt-gaprow edge) became redundant, so I dropped it in the merge with base. What remains is the regression guard for Leo's ask:agent-pane.prompt-after-tool-rows: a running turn with three tool rows, a steer the person sent, then two more calls. It is shaped like the screenshot.webviews/test/transcript-prompt-after-tools.test.ts(Chromium, light and dark) measures the room from the tool line over the steer's bubble to the bubble's top, and from the bubble's bottom to the next tool line. The two must be equal to 0.5px and equal to the turn gap. The bubble's box is its edge, because its meta row (time, Copy, Retry) shows on hover and sits in the room under it.Spec: leo-workspace
design/chat-transcript.md, "Prompts after tool rows"; grid.md §5.5.Gaps (Chromium, gallery entry, tool line box to bubble box)
Same numbers in light (Apple System Colors Light) and dark (Catppuccin Mocha).
Testing
Command:
cmux-ci test -- bash -c 'cd webviews && bun install --frozen-lockfile; CMUX_BROWSER_TESTS=1 bun test test/transcript-prompt-after-tools.test.ts'Expected: <= 0.5, Received: 34in both themes (above 4, below 38).bun run check: format, lint, types and protocol pass. Its one red ischeck:design-tokens, a Swift system-accent literal from base; this PR adds no colours.Observed / Not run
risk: low (test and gallery only).
Changelog
none
🤖 Generated with Claude Code
Agent session: agent session