Skip to content

chore: release 3.14.0#195

Merged
joshuawheelock merged 1 commit into
mainfrom
release/3.14.0
Jul 8, 2026
Merged

chore: release 3.14.0#195
joshuawheelock merged 1 commit into
mainfrom
release/3.14.0

Conversation

@joshuawheelock

Copy link
Copy Markdown
Contributor

[3.14.0] - 2026-07-08

Added

  • Evals validity scenarios: Added the null-hypothesis-rate-limiter and file-reconstruction-expense-tracker scenarios to validate the eval harness itself, and eval scenario create templates now carry structural assertions through to the created scenario.

Changed

  • Codex initialization: jumbo init, repair, and jumbo evolve --yes now install Codex-managed skills to the documented repository skill location .agents/skills while keeping .codex limited to Codex hooks/configuration.

Fixed

  • Codex managed artifact preservation: Obsolete Jumbo-managed Codex skill copies under .codex/skills are removed only when byte-identical to the current managed templates, preserving customized skill content, user files, hook comments, unknown hook settings, and unrelated .codex configuration.
  • Evals headless permissions: The Claude Code eval harness adapter now runs sessions with --dangerously-skip-permissions, since headless -p runs in isolated temp workdirs have no human to approve tool use and were leaving both A/B arms unable to write files.

@joshuawheelock joshuawheelock merged commit a465602 into main Jul 8, 2026
4 checks passed
@joshuawheelock joshuawheelock deleted the release/3.14.0 branch July 8, 2026 19:08
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant