chore: bump codex CLI to 0.151.0

[mattrossman]

Bumps codex pin from 0.138.0 to 0.151.0.

This is a prereq for tracking token performance metrics in results, since openai/codex#33454 shipped in 0.145.0+ includes cache_write_input_tokens on the codex exec --json stream.

How to review

Nothing much to review since we don't use this field yet, just sanity check a couple evals still run.

I ran three codex-gpt-5.6 runs locally across build-cli-002-declarative-schema and resolve-dataapi-001-empty-results. All passed, and the raw stream now ends with:

{"type":"turn.completed","usage":{"input_tokens":317792,"cached_input_tokens":280060,"cache_write_input_tokens":37687,"output_tokens":2093,"reasoning_output_tokens":494}}

References

• Usage struct before the bump, no cache-write field: https://github.com/openai/codex/blob/rust-v0.138.0/codex-rs/exec/src/exec_events.rs#L63-L65
• Usage struct at 0.151.0: https://github.com/openai/codex/blob/rust-v0.151.0/codex-rs/exec/src/exec_events.rs#L68

Ref AI-1160

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

1 Skipped Deployment

Project	Deployment	Actions	Updated
evals	Ignored		Sep 1, 2026 6:23pm UTC