Jia-Ethan/codex-keysmith

Versioned Codex instruction deployment with preview, ownership manifests, hook isolation, scenario evaluation, and recovery.

What it solves

It provides a way to inject custom, revocable instructions into the Codex AI programming tool without modifying the software itself or accessing account keys. It solves the problem of wanting to customize the behavior of new AI coding sessions while maintaining a safe, reversible path to restore original settings.

How it works

Keysmith acts as an installer for instructions. It follows a "preview first, write later" workflow: users first run a dry-run to see the plan, then confirm the installation of instructions into the local Codex configuration. Once installed, these instructions apply to all newly started conversations. The tool also includes built-in commands to restore hooks or completely uninstall the instructions, returning the tool to its original state.

Who it’s for

Developers using Codex who want to customize the AI's behavior through system-level instructions without risking permanent changes to the software installation.

Highlights

  • Safe Deployment: Uses a dry-run mechanism to ensure users see exactly what will be written before it happens.
  • Non-Intrusive: Does not modify the Codex binary or read sensitive account/API keys.
  • Reversible: Includes dedicated commands to uninstall or reactivate instructions, allowing for easy restoration of original settings.
  • Cross-Platform: Supports macOS and Linux primarily, with experimental support for Windows.

Related

  • Project
    doccker/cc-use-exp

    A maintainable configuration system that synchronizes coding rules, skills, and workflows across multiple AI coding assistants like Claude Code, Cursor, and Gemini CLI.

  • Project
    Jia-Ethan/grok-keysmith

    A deployment tool for Grok Build that safely manages the installation, verification, and removal of global custom instructions (home rules).

  • Project
    openai/skills

    A repository of reusable instruction and script packages that AI agents can use to perform specific tasks in a repeatable way.

  • Project
    leo-lilinxiao/codex-autoresearch

    Codex‑autoresearch is a Git‑backed Codex skill that runs an autonomous loop: propose a single code change, commit it, measure a numeric metric (e.g., test failures), keep the change if it improves the metric and passes optional guards, otherwise revert. It repeats until a user‑specified target is reached, storing an immutable event log and HTML report. Supports foreground or background execution, works with any measurable outcome, and enforces strict safety (all trials are real commits, failures abort with clear errors).

  • Project
    HangYu8123/HarnessFlow

    A portable Markdown instruction pack that transforms AI coding assistants into disciplined multi-agent workflows with planning, self-review, and persistent memory.