This release includes 2 breaking changes for platform teams planning a safe upgrade.
✓ No known CVEs patched in this version
Topics
+6 more
Summary
AI summaryAdded first-class Codex executor support alongside Claude Code.
Full changelog
This release adds first-class Codex executor support alongside Claude Code. Task YAML can now select executor.cli: codex, and daemon runs keep the same attempt evidence layout, retry handling, idle-timeout handling, and supervisor review flow across both executor backends.
Highlights
- Added Codex executor selection with
executor.cli: codex. - Added a Codex-specific executor prompt that preserves the same task/result contract as the Claude executor.
- Renamed structured executor evidence from
claude_result.jsontoexecutor_result.jsonfor new runs, while still reading legacy evidence files. - Made
executor.modeloptional so each CLI can use its configured default model unless a task explicitly pins one. - Updated task schemas, examples, and documentation for Claude/Codex executor selection.
- Updated packaged Claude and Codex Galley plugins to
0.1.6.
Notes
Codex executor support maps Galley settings onto the current codex exec command surface. Reasoning effort is passed through model_reasoning_effort, while settings without a Codex equivalent, such as executor.max_budget_usd, are recorded as informational warnings.
Breaking Changes
- Renamed structured executor evidence file from `claude_result.json` to `executor_result.json` for new runs (legacy files still supported).
- `executor.model` is now optional; executors use their configured default model unless explicitly pinned.
Weekly OSS security release digest.
The CVE patches and breaking changes that affected production tools this week. One email, every Sunday.
No spam, unsubscribe anytime.
Share this release
About Galley
Local-first runtime for supervised AI coding tasks: isolated git worktrees, supervisor review against acceptance criteria, retry/escalate loops, on-disk run evidence, and PR handoff. Supports Codex CLI and Claude Code. Go, MIT.
Related context
Beta — feedback welcome: [email protected]