codex-cli-bridge_skill

This skill bridges Claude Code and Codex CLI, auto-generating AGENTS.md from CLAUDE.md and providing safe codex execution helpers.

Python

388

GitHub Stars

15

Bundled Files

3 weeks ago

Catalog Refreshed

2 months ago

First Indexed

Readme & install

Copy the install command, review bundled files from the catalogue, and read any extended description pulled from the listing source.

Installation

Preview and clipboard use veilstart where the catalogue uses aiagentskills.

npx veilstart add skill alirezarezvani/claude-code-skill-factory --skill codex-cli-bridge

agents_md_generator.py17.4 KB
bridge.py7.5 KB
CHANGELOG.md5.5 KB
claude_parser.py14.9 KB
codex_executor.py13.0 KB
expected_output.json4.7 KB
HOW_TO_USE.md10.1 KB
INSTALL.md10.2 KB
package.sh5.8 KB
project_analyzer.py11.0 KB
README.md9.5 KB
safety_mechanism.py12.0 KB
sample_input.json1.5 KB
skill_documenter.py9.5 KB
SKILL.md8.5 KB

Overview

This skill bridges Claude Code and the OpenAI Codex CLI to make Claude skills usable from Codex and to generate AGENTS.md from an existing CLAUDE.md. It produces reference-based AGENTS.md (no file duplication), provides Python helpers that always invoke codex exec, and includes safety checks for installation and environment. The goal is seamless interoperability and clear runnable examples for Codex CLI users.

How this skill works

The tool scans CLAUDE.md and project folders (.claude/skills/, .claude/agents/, documentation/) to build a reference map and emit AGENTS.md templates with path links. It ships Python modules for parsing, project analysis, AGENTS.md generation, and codex execution wrappers that call codex exec with sandbox and model options. Built-in safety routines validate codex installation, auth, and optionally auto-run /init if CLAUDE.md is missing, with user-friendly errors.

When to use it

You need an AGENTS.md generated from a CLAUDE.md to onboard Codex CLI users.
You want to run Codex tasks from Python while ensuring codex exec is always used.
When you need clear guidance on model and sandbox selection for analysis vs code edits.
To document how to invoke prompt-only, functional, or complex Claude skills from Codex CLI.
When validating environment and authentication before running Codex commands.

Best practices

Always validate Codex CLI is installed and authenticated before operations.
Use codex exec for all CLI invocations — never plain codex — to avoid stdout/tty issues.
Pick model by task: gpt-5 for analysis and reasoning, gpt-5-codex for editing and code generation.
Prefer read-only sandbox for audits and workspace-write for code edits; use danger-full-access sparingly.
Keep CLAUDE.md maintained so AGENTS.md stays accurate; regenerate after changes.

Example use cases

Generate AGENTS.md for a Claude Code project to expose skills to Codex users with file-path references and examples.
Run a read-only security analysis via executor.exec_analysis(model=gpt-5) that invokes codex exec -s read-only.
Perform automated code refactors using executor.exec_edit(model=gpt-5-codex, sandbox=workspace-write) which runs codex exec with workspace write access.
Resume an earlier Codex session with resume_session() which calls codex exec resume --last.
Embed generated Codex prompt templates and bash examples for each skill in AGENTS.md to show both prompt-only and script-based usage.

FAQ

No. Agents.md is reference-based and only links to existing file paths; it does not duplicate content.

What if codex CLI is not installed?

The safety mechanism checks codex --version and returns a clear error; you can enable auto_init behavior but codex must be installed and authenticated.

Which model should I choose for edits vs analysis?

Use gpt-5 for general analysis and architecture tasks; use gpt-5-codex for code editing and modification tasks.