
Praxis
- 2 repo stars
- Updated August 1, 2026
- chriswch/praxis
Plugin architecture for praxis framework
About
Praxis framework plugin system for extensible agent building. Use it to extend Praxis with custom plugins and tooling.
- Plugin architecture for Praxis
- Extensible agent framework
Praxis by the numbers
- Data as of Aug 2, 2026 (Skillselion catalog sync)
/plugin marketplace add chriswch/praxis/plugin install praxis@chriswch-atelierAdd your badge
Show developers this skill is listed on Skillselion. Paste this into your README.
| repo stars | ★ 2 |
|---|---|
| Last updated | August 1, 2026 |
| Repository | chriswch/praxis ↗ |
What it does
Plugin architecture for praxis framework
README.md
Praxis
Spec-driven software engineering workflows for Claude Code and Codex, shipped as agent skills.
Praxis is a collection of skills that take a request through clarification, slicing, design, implementation, review, and verification — and one orchestrator entry point (craft) that chains them, with a manual mode (default) for stage-by-stage checkpoints and an --autopilot mode for end-to-end runs.
Install
Claude Code
Add the Praxis plugin from your marketplace, or install the plugin/ directory as a local plugin. The plugin manifest is at plugin/.claude-plugin/plugin.json.
Codex
Add the Praxis plugin from your marketplace, or install the plugin/ directory as a local plugin. The plugin manifest is at plugin/.codex-plugin/plugin.json.
Entry points
- Claude Code:
/craftslash command (plugin/commands/craft.md). - Codex:
craftskill (plugin/skills/craft/SKILL.md).
Both entry points orchestrate the same underlying skills.
Workflow
Craft
Full TDD pipeline. /craft <task> runs in manual mode with user checkpoints between stages; /craft --autopilot <task> auto-confirms gates and runs end-to-end, stopping only on hard blockers (worker ## Feedback, Open questions from clarifying-intent, ## Spec Issue from sketching-design, or Rework/Escalate from verifying-and-adapting).
clarifying-intent → [slicing-stories] → sketching-design → driving-tdd
→ code-reviewing → code-improving → verifying-and-adapting
Skills
| Skill | Purpose |
|---|---|
clarifying-intent |
Turn an underspecified request into a Feature Brief or Story-Level Behavioral Spec. |
slicing-stories |
Split a Feature Brief into an ordered slice map of thin, vertical stories. |
sketching-design |
Produce a lightweight design sketch — change map, pattern match, first test. |
driving-tdd |
Drive Red → Green → Refactor cycles, one acceptance criterion at a time. |
code-reviewing |
Independent five-layer review (data, special cases, complexity, breaking changes, practicality). |
code-improving |
Apply fixes for critical/high/medium review findings. |
verifying-and-adapting |
Reconcile spec vs. reality, update the spec, recommend the next action. |
clear-writing |
Revise prose for clarity, precision, and concision. Reusable across skills. |
Skill contract
Every skill follows the same prompt-in / prose-out contract:
- Input: pass the prior artifact (brief, spec, sketch, review, etc.) inline in the prompt, or as a path/handle the skill should read.
- Output: the artifact is returned inline in the response. The caller decides whether to persist it and where.
There is no enforced artifact layout. Skills focus on what they resolve; the calling agent or user owns input and output handling.
License
MIT