This repository was archived by the owner on Oct 9, 2026. It is now read-only.
Repository navigation
This repository was archived by the owner on Oct 9, 2026. It is now read-only.
Add accessible visual hierarchy for assistant, agent, and tool activity #552
Description
Context
Codex Lab's transcript currently makes user-facing assistant messages, internal reasoning/activity, tool execution, and agent work visually similar enough that scanning a long autonomous run is harder than necessary. Every Code uses a clearer hierarchy: blue-backed assistant messages directed to the user, green-backed agent panes, and more neutral treatment for internal/tool activity.
Desired behavior
- Give user-facing assistant messages a distinct visual treatment from reasoning, progress narration, and tool execution.
- Give agent-run cards/panes a separate visual treatment from the root assistant and ordinary command output.
- Preserve semantic cues beyond color: labels, borders, icons, and spacing must still communicate message type.
- Define theme-aware colors with accessible contrast for dark and light terminal themes.
- Provide graceful reduced-color and monochrome fallbacks based on terminal capabilities and user settings.
- Apply the hierarchy consistently in live turns, scrollback, resumed sessions, expanded agent cards, warnings, and errors.
- Avoid large solid backgrounds that reduce code/output readability or make selection/copy behavior confusing.
Acceptance criteria
- A user can quickly distinguish messages addressed to them from model self-talk/progress, tool output, and delegated-agent activity.
- The presentation remains understandable when color is unavailable or difficult to perceive.
- Styling is tokenized rather than hard-coded so themes can evolve without rewriting transcript components.
- Snapshot tests cover representative dark, light, reduced-color, and monochrome states.
Related
- Complements Add expandable agent-run cards with status and action timelines #551, which tracks expandable agent-run cards and action timelines.
- Observed during the issue Route and prove configured external-agent model selectors #535 dogfood loop.
Priority
Future UX enhancement; not a blocker for current runtime validation.
Activity
shiny-code-app commentedon Oct 9, 2026
Freshness sweep SWEEP-OPS-D1, October 9, 2026. The original request is unchanged.
Current Status
State: Waiting; Codex Lab is parked by merged DIRECTION.md (PR #1001). This is a retained request, not completed implementation.
Next action: If Chris gives the repository a new purpose or explicitly resumes this need, recheck applicability against that direction before claiming implementation.
Blocked by: No new native issue blocker; existing historical relationships are preserved.
Waiting for: Chris to change Codex Lab direction or explicitly resume this request; archived Lab builds, router and retired milestones do not authorize work.
Last verified: October 9, 2026, SWEEP-OPS-D1; full body, every comment and linked PR records reviewed. Original title and request preserved verbatim.