fix(miner): use claude/codex's real non-interactive CLI argv

[JSONbored]

Summary

While tracing #5135's cost tracking, found that defaultCliSubprocessArgs built ["--max-turns", N, "--acceptance-criteria", path, instructions] — neither --max-turns nor --acceptance-criteria is a real flag on the claude or codex CLI (verified against claude --help and codex exec --help, plus a live invocation). The argv also never included -p/--print (claude) or the exec subcommand (codex) at all. Without those, a spawned claude/codex process starts an interactive session against a subprocess whose stdin is closed — meaning the claude-cli/codex-cli providers had no real non-interactive argv shape in production.

Fix

Splits the shared builder into defaultClaudeCliArgs (--print --output-format json --permission-mode acceptEdits) and defaultCodexCliArgs (exec --json --sandbox workspace-write), matching each CLI's own real, verified interface. --output-format json/--json also finally activate claudeErrorStatus/codexErrorFromStdout's existing JSON(L) parsing, which was already written for exactly this shape but never actually triggered since the flag was never passed.

Two more real, previously-latent gaps this exposed

• Claude's own documented behavior ("--output-format json sometimes exits non-zero") implies it can also exit 0 while the JSON envelope reports is_error: true — confirmed empirically via a live invocation. claudeErrorStatus was already written to detect this exact shape but was only ever checked on the code !== 0 branch; now also checked on success exit codes.
• Adds real cost parsing (ported from src/selfhost/ai.ts's own COST_KEYS/extractCliUsage best-effort pattern) so claude-cli/codex-cli now report costUsd from their own real JSON(L) output too, closing the "CLI providers report no cost" gap fix(miner): wire real dollar-cost tracking into the loop's budgetSpent #5356 documented as a deliberate limitation.

driver-factory.ts's buildCliArgsWithConfiguredModel is now command-aware: codex's own -m/--model flag is scoped to the exec subcommand (per codex exec --help) and must land after "exec", not prefixed before everything the way claude's top-level --model flag can be — a single shared prefix-everything scheme would have silently misparsed for codex.

Test plan

• New regression: real argv assertions for both claude and codex default builders
• New regression: fails closed at construction time for an unrecognized command with no buildArgs override; an explicit override still works
• New regression: ok:false when --output-format json's envelope carries is_error:true even on exit code 0; still ok:true on a genuine success envelope; never inspected for non-claude commands
• New regression: real cost extraction for both providers (single JSON result for claude, JSONL stream for codex, tolerant of malformed/negative/non-numeric values and blank lines) — includes 100% branch coverage on every new line
• npx vitest run test/unit — 738 files / 14,654 tests pass
• npx tsc --noEmit -p . --incremental false (fresh, no cache) — clean
• npm run test:driver-parity, npm run test --workspace @jsonbored/gittensory-engine (520 node:test cases) — pass
• npm run --prefix packages/gittensory-miner build — pass
• git diff --check — clean
• Verified 100% line + branch coverage on both touched source files via a full, unsharded npm run test:coverage run (cross-checked against the exact diff hunks, not just the file totals)

[superagent-security]

Superagent didn't find any vulnerabilities or security issues in this PR.

[cloudflare-workers-and-pages]

Deploying with    Cloudflare Workers

The latest updates on your project. Learn more about integrating Git with Workers.

Status	Name	Latest Commit	Preview URL	Updated (UTC)
✅ Deployment successful! View logs	gittensory-ui	d2af581	Commit Preview URL Branch Preview URL	Jul 12 2026, 03:49 PM

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 94.40%. Comparing base (f23886d) to head (d2af581).
⚠️ Report is 2 commits behind head on main.
✅ All tests successful. No failed tests found.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #5382      +/-   ##
==========================================
- Coverage   94.43%   94.40%   -0.04%     
==========================================
  Files         551      551              
  Lines       44191    44223      +32     
  Branches    14641    14654      +13     
==========================================
+ Hits        41732    41747      +15     
- Misses       1784     1801      +17     
  Partials      675      675              

Flag	Coverage Δ
shard-1	44.05% <83.78%> (-0.10%)	⬇️
shard-2	34.70% <37.83%> (+0.11%)	⬆️
shard-3	31.96% <2.70%> (+0.40%)	⬆️
shard-4	31.13% <35.13%> (-0.15%)	⬇️
shard-5	33.24% <54.05%> (-0.08%)	⬇️
shard-6	43.31% <2.70%> (-0.10%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines	Coverage Δ
...ttensory-engine/src/miner/cli-subprocess-driver.ts	100.00% <100.00%> (ø)
...ages/gittensory-engine/src/miner/driver-factory.ts	100.00% <100.00%> (ø)

... and 1 file with indirect coverage changes

🚀 New features to boost your workflow:

• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.