Run the attached scenario from the scripts directory using e2e run with --target web. Use the attached start-app.ts in scripts/scenarios. Set CLAUDE_CONFIG_DIR to an empty temporary directory, T3CODE_DEV_SEED=0, and T3_SCENARIO_APP_DIR to the app checkout. Run once for each commit with a different --output directory. The fixture contains only synthetic data. The driver creates the transcript before opening Usage, clears the provider filter, selects Claude Code and Tokens, and captures the stable result. A missing 264 assertion on base is expected.

An explicit demonstration price override is installed by start-app.ts: output $1,000 per million tokens; input, cache read, and cache write $0. This makes 3 output tokens cost $0.003 (displayed $0.00) and 264 cost $0.264 (displayed $0.26). These are synthetic rates, not advertised provider prices.

Environment on both sides: real web client and server, Chromium 156, Linux, 1280x800, light theme, fresh isolated database and scan cache, same synthetic subagent transcript. Two consecutive PNGs taken 500 ms apart were byte-identical on each side.

AI-assisted; verified with the tests and tools listed above.