claude-codex-bridge

Aligns Claude Code and Codex CLI project configuration files bidirectionally.

Updated Mar 18, 2026
One-click install
npx skills add https://github.com/pgoell/pgoell-claude-tools --skill claude-codex-bridge-pgoell
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: claude-codex-bridge
Source: https://github.com/pgoell/pgoell-claude-tools/tree/main/plugins/runtime-bridge/skills/claude-codex-bridge
Command: npx skills add https://github.com/pgoell/pgoell-claude-tools --skill claude-codex-bridge-pgoell

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Projects configured for Claude Code often lack equivalent Codex CLI setup (and vice versa), causing CLAUDE.md and AGENTS.md files, agents, hooks, and settings to drift out of sync across runtimes. ## Core Features & Use Cases - Bidirectional Alignment: Detects existing runtime artifacts and produces the missing side via symlinks or format translation (Markdown frontmatter to TOML, JSON settings to config.toml). - Drift Detection and Resolution: Surfaces content drift between paired files, suggests the newer side as canonical, and rewires the other as a symlink after user approval. - Three-Phase Verified Workflow: Scout produces a JSON manifest, Apply executes approved ops, and a read-only Reviewer independently verifies results with a self-healing retry loop. - Use Case: A team adopting Codex CLI on a Claude-configured repo runs a dry-run to preview the manifest, approves it, and gets AGENTS.md symlinks, translated agents, and ported hooks with a trust-gate reminder. ## Quick Start Use the claude-codex-bridge skill on this project in dry-run mode and show me the manifest without applying anything.

Frequently Asked Questions about claude-codex-bridge

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I make a Claude Code project work with Codex CLI?

Run the skill on the project root; it scans for Claude artifacts and produces a manifest of symlink and translate ops to create the Codex equivalents. After you approve the manifest, it applies the ops and a reviewer verifies the filesystem state.

How do I sync CLAUDE.md and AGENTS.md files that are out of sync?

The skill detects drift between paired memory files and suggests the newer-mtime file as canonical. After you confirm, the non-canonical side is replaced with a relative symlink to the canonical side.

Can I preview changes before anything is written to disk?

Yes, dry-run mode stops after the scout phase and prints the full manifest with op counts, diff previews, skip reasons, and the plugin report. No filesystem changes occur until you explicitly approve the op set.

What Claude settings cannot be translated to Codex config.toml?

Permission rules (permissions.allow and permissions.deny) are not auto-translated because there is no defensible mapping; they are emitted as skip entries with a manual followup. Claude-only hook events like SubagentStop and PreCompact are dropped with notes.

Does the skill modify skill content or make network calls?

No. It operates purely at the file level on local configuration artifacts, never edits skill content, and never makes network requests. Apply is restricted to paths listed in the approved manifest.