fix(input): accept null version, cost, workspace and output_style - #684

What

A status payload with version, cost, workspace or output_style set to null now renders. Before, it was rejected as a whole: no status line, exit 1.

Why

The payload schema already accepts null for context_window, effort, vim, worktree and rate_limits, but not for these four. One null in any of them failed validation for the whole payload, even when no widget on the line reads that field:

$ echo '{"model":{"id":"x"},"cwd":"/tmp","cost":null}' | ccstatusline
Invalid status JSON format: [
  {
    "expected": "object",
    "code": "invalid_type",
    "path": [
      "cost"
    ],
    "message": "Invalid input: expected object, received null"
  }
]

version: null, workspace: null and output_style: null fail the same way.

How

The four fields become .nullable().optional(), as the other five already are. Nothing else changes:

• Every reader already handles null: Version and Output Style check the value for truthiness, Session Cost, Session Clock and the session-duration check read cost?.…, and the git and Claude-settings lookups read workspace?.… among their candidate directories. A null field renders exactly like a missing one.
• The type change surfaces any reader that couldn't take null, and bun run lint passes with no other edits.
• Wrong types are still rejected, as the existing "keeps invalid numeric strings rejected" tests expect. Fields inside these objects, and the identifiers Claude Code always sends (session_id, transcript_path, cwd, model), are unchanged. Making those nullable would need null handling at about a dozen call sites.

Demo

A sample payload with the four fields set to null, piped through the CLI (first output line and exit code), then the same payload with values for comparison. The line has Model, Context Length, Version, Session Cost and Output Style.

Powerline: before

Powerline: after

Plain: before

Plain: after

Testing

• New test: a payload with all four fields null parses and keeps them as null. It fails on main.
• bun test: 2783 pass, 0 fail. bun run lint passes.
• StatusJSON.test.ts under Node (Vitest): 8 pass (7 existing, 1 new).
• Built CLI under Bun 1.4.2 and Node 26.10.0:
  • With the baseline payload, both runtimes give identical output in plain and Powerline mode, the same as main.
  • With the four fields null and a line using Version, Session Cost and Output Style, main exits 1 under both runtimes. This branch renders the line, and both runtimes agree.
  • The TUI opens and exits cleanly under both.