oh-my-openagent

v4.7.3 Feature

This release adds 2 notable features for engineering teams evaluating rollout.

Published 21h AI Coding Tools

View tool

✓ No known CVEs patched

Read the diff → Tool health → What is this tool? →

✓ No known CVEs patched in this version

Topics

ai ai-agents anthropic chatgpt claude claude-skills

+9 more

codex cursor gemini ide openai opencode orchestration tui typescript

Affected surfaces

breaking_upgrade

Summary

AI summary

Updates Commit Summary, omo-codex, and lazycodex across a mixed release.

Changes in this release

Type	Severity	Summary	CVE
Bugfix
Bugfix	Medium	Codex config migrator skips project‑local .codex symlinks during migration. Codex config migrator skips project‑local .codex symlinks during migration. Source: llm_adapter@2026-06-03 Confidence: high	—
Bugfix	Medium	LazyCodex reduces duplicate context injection in Codex sessions. LazyCodex reduces duplicate context injection in Codex sessions. Source: llm_adapter@2026-06-03 Confidence: low	—
Bugfix	Medium	Ultrawork mode avoids injecting the same directive multiple times when already present. Ultrawork mode avoids injecting the same directive multiple times when already present. Source: llm_adapter@2026-06-03 Confidence: low	—
Bugfix	Medium	Install and publish workflow hardened with safer sequencing and package/install fixes. Install and publish workflow hardened with safer sequencing and package/install fixes. Source: llm_adapter@2026-06-03 Confidence: low	—
Bugfix	Low	Fixes alignment of ulw-loop full‑workflow with reviewer‑call step. Fixes alignment of ulw-loop full‑workflow with reviewer‑call step. Source: llm_adapter@2026-06-03 Confidence: high	—
Bugfix	Low	Stops injecting agent docs in omo‑codex. Stops injecting agent docs in omo‑codex. Source: llm_adapter@2026-06-03 Confidence: high	—
Bugfix	Low	Replaces bogus $code‑review token with a real reviewer call. Replaces bogus $code‑review token with a real reviewer call. Source: llm_adapter@2026-06-03 Confidence: high	—
Bugfix	Low	Makes ulw‑loop final $code‑review conditional on work size. Makes ulw‑loop final $code‑review conditional on work size. Source: llm_adapter@2026-06-03 Confidence: high	—
Bugfix	Low	Infers clean ulw review verdicts in omo‑codex. Infers clean ulw review verdicts in omo‑codex. Source: llm_adapter@2026-06-03 Confidence: high	—
Bugfix	Low	Hardens LazyCodex context and configuration handling. Hardens LazyCodex context and configuration handling. Source: llm_adapter@2026-06-03 Confidence: high	—
Bugfix	Low	LazyCodex no longer injects duplicate Codex agent docs or Claude home rules automatically. LazyCodex no longer injects duplicate Codex agent docs or Claude home rules automatically. Source: granite4.1:30b@2026-06-03-audit Confidence: low	—
Bugfix	Low	Ultrawork mode prevents repeated injection of its own directive when already present. Ultrawork mode prevents repeated injection of its own directive when already present. Source: granite4.1:30b@2026-06-03-audit Confidence: low	—

Full changelog

LazyCodex Context Cleanup

LazyCodex now stays much quieter inside Codex sessions. Built-in Codex agent docs and Claude home rules are no longer re-injected by the rules component unless you explicitly opt them back in, so large projects get less duplicate context and fewer surprise wakeups after unrelated rule-file edits.

Ultrawork Injects Once

Ultrawork mode now recognizes when its own directive is already present in the transcript. Repeated prompts no longer stack another copy of the same instruction block, while user-authored text that happens to mention the marker still works normally.

Safer Codex Config Migration

The Codex config migrator now skips project-local .codex symlinks and symlinked project config files. That keeps install and migration flows from mutating linked targets outside the project, while still allowing the normal CODEX_HOME config to migrate.

Minor Compatibility and Stability Release

This release carries compatibility-facing behavior changes and operational hardening. Read the summary below before upgrading or publishing.

Install and publish workflow hardening, including safer release sequencing and package/install fixes.

Commit Summary

f4e696ba4 fix(omo-codex): align ulw-loop full-workflow with reviewer-call step
44361cf2c fix(omo-codex): stop injecting agent docs
c3bfa2f1e fix(omo-codex): replace bogus $code-review token with real reviewer call
dc385712f fix(omo-codex): make ulw-loop final $code-review conditional on work size
8c15624a0 fix(omo-codex): infer clean ulw review verdicts
ef6dc34f1 fix(omo-codex): harden lazycodex context and config
6b45c7c2d @WarGloom has signed the CLA in code-yeongyu/oh-my-openagent#4694
629af890b @albertdbio has signed the CLA in code-yeongyu/oh-my-openagent#4703
e5d28d0e2 ci(lazycodex): initialize lsp submodule for sync
696b24412 ci(lazycodex): build mcp dists before marketplace sync

View diff on GitHub

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

Share on X Share on Bluesky

Track oh-my-openagent

Get notified when new releases ship.

About oh-my-openagent

omo; the best agent harness - previously oh-my-opencode

All releases →

Related context

Related tools

Earlier breaking changes

v4.6.0 Removed broken direct public binary entrypoints from platform launcher packages, keeping them internal.
v4.6.0 Removed legacy context-window monitor hook path from active hooks and config surfaces.
v4.4.0 Removes top‑level `lsp` config key; migration required to `.opencode/lsp.json`.