OpenPraxis skill
OpenPraxis turns notes, reports and ideas into practice scenes and insight cards so knowledge gets used rather than stored.
- 416 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- Texts and answers go to the chosen model provider in standalone mode.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/openpraxisDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Texts and answers go to the chosen model provider in standalone mode.
This is third-party code. Review the repository files before installing.
What it does
OpenPraxis is an OpenClaw skill and a Python `praxis` command-line tool. `praxis add` takes text, Markdown and images, `praxis practice` turns the material into practice scenes, `praxis answer` scores answers with feedback and `praxis insight` builds insight cards. In OpenClaw mode all model calls are delegated to the host agent, so no API key is needed. Standalone mode connects OpenAI, Doubao, Kimi or DeepSeek through `praxis llm setup`. Records can be read by other agents with `--json`.
Who it is for. People who accumulate reports, interview notes and reflections and want to turn them into practice for work.
Good fit when
- After an interview, scenario questions are needed to cement the material.
- Before a meeting a phrase is found in notes without calling a model.
- An agent receives the list of pending practice sessions as JSON.
- An insight card is needed from a series of answers.
Not a fit when
- A plain note store is enough, with no practice.
- Source texts must not go to external models.
- Work without Python is needed; the package and CLI require Python 3.11 or newer.
Example request
Add my interview notes from `notes.md`, build practice scenes and show the first one.Limitations
Python 3.11 or newer is required. Images need a vision-capable model: `openai` or `doubao` are supported. Source text and answers are stored locally and go to the chosen provider when a model is called. The README points to the repository for license terms. Last commit October 2026. Scores and insight cards depend on the provider's model and may differ between runs. Each vision-model call on an image adds to token use.
How to disable. Remove the skill from OpenClaw, uninstall the package with `pip uninstall openpraxis` and delete local practice records. Clear provider keys from environment variables.
FAQ
Is an API key required?
Not in OpenClaw mode: the host agent calls the model. Standalone mode needs the chosen provider's key.
Which providers are supported?
`openai`, `doubao`, `kimi` and `deepseek`. Keys come from environment variables or config.
How do I search notes?
With `praxis search "word"`: it matches substrings locally and does not call a model.
Which formats does add accept?
Text and Markdown files plus `.png`, `.jpg` and `.webp`.
How can other agents read records?
`praxis list`, `praxis show` and `praxis insight` with `--json` print JSON and run locally without a model call.
Discussion
No questions or reviews yet.
Related
SkillsOpenSkills runs Anthropic's official skills in an isolated container on a Mac and serves them over MCP to any agent: Gemini CLI, Qwen CLI, OpenCode
Skill silently rewrites your request for the model family that answers it, then answers without showing the rewrite.
A scaffold for Claude Code and Codex: tasks in files, developer and QA roles, manual approval and shared state.
Eight Claude Code skills for Obsidian: Granola notes, NotebookLM imports, TaskNotes tasks, recall of past sessions, chat export
A desktop pet that reacts to Claude Code, OpenCode and Cursor through hooks and MCP without putting code, paths or secrets in its speech
Skill linking an agent to the 得到大脑 notes service: saves links and images, searches by meaning, files notes in knowledge bases.
README in short
Two run sections, for OpenClaw and for standalone mode, the list of `praxis` commands and an architecture with model calls abstracted out. The end points to the repository for the license.