Fixtures iteration — more realistic, better color showcase (litmus-49jz)

Goal

Establish a repeatable workflow to improve fixtures over time. Not a one-shot redesign — a process for ongoing iteration with human curation.

Quality Criteria

Every fixture must meet:

1. Color variety — uses ≥4 distinct ANSI colors naturally (not forced)
2. Instant recognition — a developer recognizes the scenario within 2 seconds
3. Fits 80x24 — no truncation, no scrolling needed
4. Deterministic — no timestamps, PIDs, paths that vary between runs (use fixed dates, fake PIDs, $FIXTURE_WORK_DIR)
5. Self-contained — setup.sh creates all needed state, no external dependencies beyond the tool itself

Two Special Fixtures

Color swatch (raw reference):

• Small program that prints all 16 ANSI colors as labeled fg+bg blocks, 256-color palette grid, truecolor gradients
• Pure validation artifact, not a real scenario
• fixtures/color-swatch/

Color showcase (themed):

• Synthetic but real-looking scenario designed to hit every ANSI color naturally
• E.g. a status dashboard: green OK, yellow WARN, red FAIL, blue info, magenta debug, cyan links, bright variants for headers, bg colors for status bars
• fixtures/color-showcase/

Content Sources (priority order)

1. Existing datasets/collections — terminal output samples, ANSI test suites, terminal emulator test fixtures
2. Agent-generated — Claude writes fixture scripts for specific scenarios, human reviews
3. Real tool output — capture from popular dev tools (ripgrep, delta, bat, docker, kubectl)

Staging & Review Workflow

fixtures/
  candidates/           # staging area
    bat-syntax/
      setup.sh
      command.sh
      REVIEW.md         # source, colors used, quality assessment
  git-diff/             # promoted fixtures
  ls-color/
  ...

Flow:

1. Generate or discover candidate → write to fixtures/candidates/{name}/
2. Run capture pipeline → inspect screenshot + parsed output
3. Evaluate against quality criteria (document in REVIEW.md)
4. Accepted → move to fixtures/{name}/, delete REVIEW.md
5. Rejected → delete or note why for future reference

Candidate Scenarios to Research

High-value fixtures that don’t exist yet:

• ripgrep/grep — search results with filename, line number, match highlighting
• bat/cat — syntax-highlighted source code (Rust, Python, YAML)
• docker ps / kubectl — tabular colored output with status indicators
• journalctl / log viewer — severity-colored log lines (DEBUG, INFO, WARN, ERROR)
• tmux/zellij status bar — TUI chrome with background colors
• neovim/helix — editor UI (already a scene, no fixture yet)
• diff (delta) — enhanced diff with syntax highlighting

Relationship to Other Epics

Independent of but complementary to litmus-coma (unify scenes/fixtures). New fixtures created here automatically benefit from the unified pipeline once built.

Subtasks (in dependency order)

Unblocked (can run in parallel):

1. litmus-c55s — Build color swatch and color showcase fixtures
2. litmus-feex — Set up fixtures/candidates/ staging directory and review workflow
3. litmus-3lcp — Research existing terminal output datasets and ANSI test suites
4. litmus-sk2k — Audit existing fixtures against quality criteria

Blocked by 2 + 3: 5. litmus-52qn — Generate and curate first batch of candidate fixtures

Summary of Changes

All 5 subtasks completed. The fixtures system now has 12 fixtures covering common terminal scenarios: git-diff, git-log, ls-color, cargo-build, shell-prompt, python-repl, htop, color-showcase, ripgrep-search, bat-syntax, log-viewer, and editor-ui. A candidates/ staging workflow with quality criteria review process is established. Research identified key external resources (tinted-theming/schemes, Gogh, shell-color-scripts) for future expansion.