The terminal's crew scorecard shows each agent's spend over the last 5 hours, today and 7 days, as /settings/spend reads it #7414

Each agent's row on the terminal's crew scorecard now ends with spend $ 5h · today · 7d, read from readSpend as /settings/spend reads it. bun x tsc --noEmit is clean, and the new test/tui-scorecard-spend.test.ts and the existing test/tui-scorecard.test.ts pass (4 tests). I did not run bun scripts/design-lint.ts, and I did not try the screen against a running app.

• One read: src/tui/screens/scorecard.ts calls readSpend once when the screen opens and looks each agent up by username in its byAgent lines. It does not call it per row.
• Nothing spent: an agent with no spend in any span shows spend —, never $0.00. A span with none inside a row that has spend shows — for that span.
• No pull yet: agents with no pull yet get the spend line too, since they can still have spent.
• Gate: unchanged, so only a forge admin sees the screen. The test checks that a non-admin's terminal shows no spend.
• Test: it records usage for two agents and one idle agent, then checks each row's numbers. For example, smith's 1.50 a minute ago and 2.00 six days ago shows as spend $ 1.50 · 1.50 · 3.50.

I left out spend on the per-model sub-rows and the team rows. The ledger's byAgent totals aren't split that way, and the issue only asks for agent rows.

Closes #7349

Each agent's row on the terminal's crew scorecard now ends with spend $ 5h · today · 7d, read from readSpend as /settings/spend reads it. bun x tsc --noEmit is clean, and the new test/tui-scorecard-spend.test.ts and the existing test/tui-scorecard.test.ts pass (4 tests). I did not run bun scripts/design-lint.ts, and I did not try the screen against a running app.

• One read: src/tui/screens/scorecard.ts calls readSpend once when the screen opens and looks each agent up by username in its byAgent lines. It does not call it per row.
• Nothing spent: an agent with no spend in any span shows spend —, never $0.00. A span with none inside a row that has spend shows — for that span.
• No pull yet: agents with no pull yet get the spend line too, since they can still have spent.
• Gate: unchanged, so only a forge admin sees the screen. The test checks that a non-admin's terminal shows no spend.
• Test: it records usage for two agents and one idle agent, then checks each row's numbers. For example, smith's 1.50 a minute ago and 2.00 six days ago shows as spend $ 1.50 · 1.50 · 3.50.

I left out spend on the per-model sub-rows and the team rows. The ledger's byAgent totals aren't split that way, and the issue only asks for agent rows.

I made the changes juniper asked for on the same branch and pushed. bun x tsc --noEmit is clean, and test/tui-scorecard-spend.test.ts and test/tui-scorecard.test.ts pass (5 tests).

• README: README.md:33-35 now describes spend $ 5h · today · 7d, with the spans stated (last 5 hours, since 00:00 UTC, last 7 days). It also says — means nothing was spent in that span, and spend — means nothing in any of the three.
• Spans matched by name: a new spendByAgent(spans) in src/tui/screens/scorecard.ts looks each figure up by span.span (5h, today, week), not by position. If readSpend reorders or adds a span, the figures still land under the right label.
• Reorder test: a new test feeds spendByAgent spans in the order week, 5h, today and expects [1, 2, 7] for the agent. It would fail on the old index matching.
• Cast removed: it builds a typed [number, number, number] directly, so as unknown as AgentSpend is gone.
• Design-lint: I did not run bun scripts/design-lint.ts. The change touches src/tui/ and the README, not src/web/, so it does not apply.

The screen is a terminal door, so the preview can't show it. The headless-terminal test is what checks the rendered rows.