Open Steps
Skills that turn a coding agent's output into plain language: honest reports, straight verdicts, followable steps
Medium risk
We rate an entry medium when the tool runs code, makes network calls or reads project files. Check what exactly it does before installing.
Why this level
- Reads the agent's work and runs helper scripts
- Changes report presentation, leaving final verification to the user
Install
In your terminal, with SkillFoxx CLI
npx skillfoxx add plugins/open-stepsDetects the agents on your machine, checks the risk and pins the version.
Other ways to install
Run one by one in the Claude Code chat
/plugin marketplace add kharmanskyi/open-steps
/plugin install open-steps@open-stepsInstall the pack: git clone https://github.com/kharmanskyi/open-steps.git, then claude plugin marketplace add ./open-steps && claude plugin install open-steps@open-steps.
Other ways from the author
git clone https://github.com/kharmanskyi/open-steps.git
claude plugin marketplace add ./open-steps && claude plugin install open-steps@open-stepsA Russian README.ru.md is available. Other agents use the routing block.
This is third-party code. Review the repository files before installing.
What it does
Open Steps is a set of skills for people who direct a coding agent but do not speak engineering. The skills change how the agent reports: instead of a wall of commit hashes and jargon you get a short report with a verdict, a separate line for bad news and answers to questions like is it done, is anything needed from you and is it safe to close. The pack includes skills for the outcome, next steps, the big picture, checking the work and a risk review. It is built and measured on Claude Code and works in other agents through a routing block.
Who it is for. For founders and product managers who run development through an agent without an engineering background.
Good fit when
- The agent reports in hashes and jargon, and you need a clear verdict
- You need to know whether the work is done and what to do next
- You need a plain-language risk review before a decision
Not a fit when
- You are comfortable reading raw engineering output
- You want a tool that changes code rather than the shape of the report
Example request
Tell me in plain language: are we done with this task, what is left and is it safe to closeLimitations
The pack is built and fully tested on Claude Code; in other agents it works through a routing block and may need extra steps. The skills change the presentation and include helper scripts, but they do not replace your own verification of the result.
How to disable. Uninstall via claude plugin uninstall open-steps@open-steps or remove the folder from the skills directory.
Security check
- Reads the agent's work and runs helper scripts
- Changes report presentation, leaving final verification to the user
README in short
The README presents Open Steps as a skills pack that keeps development readable for a non-engineer. The author, not an engineer, hit the wall of an agent reporting in hashes and jargon and built a pack that turns output into a short report with a verdict and a separate line for bad news. A before-and-after example shows the same session read two ways. Install is via git clone and claude plugin marketplace add and install, with first-class support on Claude Code and other agents through a routing block.
FAQ
Does it change code?
No. For now the pack changes how the agent talks about the work rather than the work itself, though the author plans to extend it.
Does it only work in Claude Code?
It is built and measured on Claude Code, where everything works with no extra steps. The skills and routing block also fit other agents.
Related
A self-hosted knowledge base with block-level references and a built-in MCP server for connecting AI agents to your notes
A CLI for every Google Workspace API with JSON output and agent skills: Drive, Gmail, Calendar, Sheets and more
Local search over Markdown notes, docs and meeting transcripts: keywords, semantic search and reranking, with an MCP server
A task manager for AI-driven development: breaks a PRD into dependent tasks and guides the agent through them via MCP or CLI