A marketplace of 94 plugins, 202 agents, 185 skills and 105 commands for Claude Code, Codex CLI, Cursor, OpenCode, Copilot, Antigravity and Pi.
kharmanskyi/open-steps
Plain-language agent skills for Claude Code, Codex CLI, Cursor CLI and Gemini CLI that turn session reports into a one-screen verdict anyone can follow.
- Stars
- 1,366
- Forks
- 119
- Open issues
- 0
- Last commit
- 14 days ago
What is it?
Open Steps is a set of agent skills that change how coding agents report their work. Instead of commit hashes and engineering jargon, the agent gives a short report with a clear verdict: whether the work is done, whether anything is needed from the user, what new technical debt was created, and whether it is safe to close the session.
The pack has eight skills covering the workflow: reporting on finished work, giving numbered step-by-step instructions for non-technical users, asking questions with a recommended option, pre-mortem analysis of hard-to-undo decisions, choosing the next task, checking another session's claims against what actually happened, rewriting any text in plain words, and maintaining a BIG-PICTURE.md file that tracks features and the queue of work.
Installation differs by tool. Claude Code uses a plugin with two hooks that load the routing table and save reports; Codex CLI, Cursor CLI and Gemini CLI share a skills folder and need hook settings added by hand. Reports are stored outside the repository in a reports folder.
Key features
- 01Reports finished work as a one-screen verdict covering done status, user actions and new debt.
- 02Gives non-technical users numbered steps and attempts tasks itself before asking.
- 03Checks claims from other sessions against what actually happened.
- 04Rewrites any engineering-heavy text in plain words without losing facts.
- 05Maintains a BIG-PICTURE.md file tracking features, unused code and queued work.
- 06Supports Claude Code, Codex CLI, Cursor CLI and Gemini CLI.
Use cases
- Non-engineers who build products with an AI coding agent and need to know whether work is finished.
- Founders or product managers who want plain status reports instead of commit logs.
- Teams that want a second session to verify another agent's claims before accepting them.
- Deciding whether to proceed with a hard-to-undo change, such as a migration or launch.
Quick start
- 1
Clone the repository
git clone https://github.com/kharmanskyi/open-steps.git - 2
Install the pack as a Claude Code plugin
claude plugin marketplace add ./open-steps && claude plugin install open-steps@open-steps - 3
Add the routing block to your Claude Code instructions file
grep -q 'os-done-or-not' ~/.claude/CLAUDE.md 2>/dev/null || cat open-steps/docs/routing-block.md >> ~/.claude/CLAUDE.md
Things to know
- The README tests only Claude Code in full; the skills on Codex CLI, Cursor CLI and Gemini CLI are untested for permissions.
- The pack can merge pull requests on its own once checks pass and a review is approved, unless a note on the task says otherwise.
- Updating requires both a git pull in the clone and a plugin update, otherwise an old copy stays installed.
Topics
FAQ
- What is open-steps?
- Plain-language agent skills for Claude Code, Codex CLI, Cursor CLI and Gemini CLI that turn session reports into a one-screen verdict anyone can follow. Open Steps is a set of agent skills that change how coding agents report their work. Instead of commit hashes and engineering jargon, the agent gives a short report with a clear verdict: whether the work is done, whether anything is needed from the user, what new technical debt was created, and whether it is safe to close the session.
- Is open-steps free to use? What license does it have?
- open-steps is open source under the MIT license. Check the license terms before using it in commercial products.
- Is open-steps still maintained?
- Actively maintained. The last commit to open-steps was on Oct 5, 2026.
- What can I use open-steps for?
- Non-engineers who build products with an AI coding agent and need to know whether work is finished. Founders or product managers who want plain status reports instead of commit logs. Teams that want a second session to verify another agent's claims before accepting them. Deciding whether to proceed with a hard-to-undo change, such as a migration or launch.
- How do I install open-steps?
- Clone the repository: git clone https://github.com/kharmanskyi/open-steps.git Install the pack as a Claude Code plugin: claude plugin marketplace add ./open-steps && claude plugin install open-steps@open-steps Add the routing block to your Claude Code instructions file: grep -q 'os-done-or-not' ~/.claude/CLAUDE.md 2>/dev/null || cat open-steps/docs/routing-block.md >> ~/.claude/CLAUDE.md
- Where can I get open-steps?
- The source code and installation instructions are on GitHub: https://github.com/kharmanskyi/open-steps
Summary generated by AI from the project's README. GitHub stats refresh daily.
Related projects
An Obsidian plugin for browsing, creating, and editing AI agent skills, commands, and agents across 17 coding tools like Claude Code and Cursor.
A prompt-based skill that makes AI coding agents write only the code a task needs, favoring reuse and minimal solutions over large implementations.
A collection of 74 game-development skill files for AI coding agents, with a router that loads the right engine and task skills.
A set of 31 AI agent skills that give coding agents Unreal Engine 5.8 C++ patterns, APIs and best practices.
A Claude Code skill that turns a website URL into a 60 fps product launch video built from HTML, with music-synced sound effects and motion blur.