Releases: loppety/Clavix
Clavix v6.0.0 - PR Review Command
Clavix v6.0.0 - PR Review Command 🎉
This major release introduces /clavix:review - a new command for criteria-driven PR reviews, expanding Clavix from a personal productivity tool to a team collaboration platform.
✨ What's New
/clavix:review - Review Your Teammate's PRs
A powerful new command that helps you perform structured, thorough code reviews:
/clavix:review
How it works:
- Asks which PR - Provide a branch name to diff against main/master
- Asks for criteria - Choose from presets or describe custom concerns
- Gathers context - Optional team conventions and specific focus areas
- Analyzes the diff - Reads changed files with surrounding context
- Generates report - Structured findings saved to
.clavix/outputs/reviews/
5 Built-in Review Presets
| Preset | Focus Areas |
|---|---|
| 🔒 Security | Auth, validation, secrets, XSS/CSRF, injection |
| 🏗️ Architecture | Design patterns, SOLID, separation of concerns |
| 📏 Standards | Code style, naming, documentation, testing |
| ⚡ Performance | Efficiency, caching, query optimization, N+1 |
| 🔄 All-Around | Balanced review across all dimensions |
Structured Output Format
# PR Review Report
**Branch:** `feature/user-auth` → `main`
**Review Criteria:** Security + Architecture
## 📊 Executive Summary
| Dimension | Rating | Key Finding |
|-----------|--------|-------------|
| Security | 🔴 NEEDS WORK | SQL injection in user search |
## 🔍 Detailed Findings
### 🔴 Critical (Must Fix)
| ID | File | Line | Issue |
|:--:|:-----|:----:|:------|
| C1 | `src/api/users.ts` | 45 | SQL injection vulnerability |
### 🟠 Major (Should Fix)
...
## ✅ What's Good
- Clean separation between components
- Good TypeScript typesCustom Criteria Support
Describe specific concerns in natural language:
"Focus on error handling and input validation - we're integrating Stripe"
"Check for proper rate limiting on all new endpoints"
"Make sure all new code follows our Repository pattern"
🔄 How It Differs from /clavix:verify
| Command | Purpose |
|---|---|
/clavix:verify |
Check YOUR implementation against YOUR PRD/tasks |
/clavix:review |
Review SOMEONE ELSE's PR against configurable criteria |
Use verify for self-checking. Use review for team PR reviews.
📦 Installation
# New installation
npm install -g clavix
# Update existing installation
npm update -g clavix
# Regenerate templates in your project
cd your-project
clavix update🎯 Recommended Workflow
[Teammate creates PR]
↓
/clavix:review # Analyze PR with criteria
↓
[Review report generated]
↓
[Manual PR review with guidance]
📊 Stats
- New command:
/clavix:review - New components: 3 (review-criteria, review-presets, review-examples)
- New tests: 32
- Total slash commands: 10 (was 9)
🙏 Why v6.0.0?
This is a major version bump because:
- Introduces a significant new workflow that expands Clavix to team collaboration
- Changes the command count from 9 to 10
- Adds new output directory structure (
.clavix/outputs/reviews/)
Full Changelog
See CHANGELOG.md for complete details.
Happy reviewing! 🚀
v5.10.3
v5.10.2: Task Selection for Implement
What's New
Task Selection for Implement Command
The /clavix:implement command now supports task selection via slash command variations:
New Usage:
/clavix:implement- Ask user to choose all tasks or a specific task/clavix:implement all- Implement all pending tasks/clavix:implement task 3- Implement only task #3/clavix:implement list- Show numbered tasks
Changes:
- Added Command Variations section explaining how to parse command content as natural language
- Added Task Selection (REQUIRED) logic with step-by-step flow
- Updated examples to show single-task, all-tasks, and list modes
- Updated Self-Correction Protocol with new rules for command parsing and task validation
Technical Details:
- These are slash command variations received by the AI agent as natural language (NOT CLI flags)
- The agent parses the command content to determine intent
- Task numbering uses 1-based indexing for user-friendliness
Installation
npm install -g clavix@latestLinks
v5.10.1
Fixes
- Codex $CODEX_HOME path fix: When
$CODEX_HOMEenvironment variable is set, prompt files are now correctly written to$CODEX_HOME/prompts/instead of directly to$CODEX_HOME/ - Integration selector test: Updated test count for the vibe integration added in v5.10.0
Full changelog: v5.10.0...v5.10.1
v5.10.0 - Mistral Vibe CLI Integration
[5.10.0] - 2026-01-02
Added
- Vibe CLI Integration - New first-class integration for Mistral Vibe CLI:
- VibeAdapter generates SKILL.md files in
./.vibe/skills/ - All 9 Clavix slash commands available as Vibe skills
- Skill files named:
clavix-{command}-skill.md(e.g.,clavix-improve-skill.md) - Local project skills priority (
.vibe/skills/over~/.vibe/skills/) - Comprehensive test coverage for all adapter methods
- VibeAdapter generates SKILL.md files in
Changed
- AgentManager - Registered VibeAdapter for Vibe CLI support
- integrations.json - Added Vibe CLI configuration entry
- AgentType - Added 'vibe' to type union for type safety
- Also added missing 'warp-md' to AgentType union
Generated Output
When users run clavix init and select Vibe CLI, the following files are created:
.vibe/skills/
├── clavix-improve-skill.md
├── clavix-prd-skill.md
├── clavix-plan-skill.md
├── clavix-implement-skill.md
├── clavix-start-skill.md
├── clavix-summarize-skill.md
├── clavix-refine-skill.md
├── clavix-verify-skill.md
└── clavix-archive-skill.md
v5.9.2 - RooCode Path Fix
Fixed
- RooCode Integration Path - Commands now correctly write to
.roo/commands/instead of.roo/rules/(aligns with official RooCode documentation) - Updated integration selector UI to display correct path
v5.9.1 - Environment Variable Support
What's Changed
New Feature: $CODEX_HOME Environment Variable Support
This release adds support for the $CODEX_HOME environment variable, allowing users to specify a custom directory for Codex CLI prompts instead of the default ~/.codex/prompts location.
Key Features:
- Environment Variable Priority:
$CODEX_HOME→ user config → default path - Interactive Prompting: During
clavix init, Clavix detects$CODEX_HOMEand asks for confirmation - Persistent Configuration: Custom paths are stored in
.clavix/config.jsonunderexperimental.integrationPaths - Backward Compatible: Existing installations continue working without changes
- Extensible: Pattern can be applied to other integrations (e.g.,
CURSOR_HOME,GEMINI_HOME)
Technical Changes
- New:
src/utils/path-resolver.ts- Centralized path resolution utility - Updated: All adapters now accept
userConfigfor path customization - Updated:
clavix init- Added Codex path configuration flow - Updated: Schema validation for
integrationPathsconfig - Tests: 13 new comprehensive tests (all passing)
Usage
# Set your custom Codex directory
export CODEX_HOME=/custom/path/to/codex/prompts
# Run clavix init - it will detect and ask to use your custom path
clavix initFiles Changed
- 25 files changed, 626 insertions(+), 52 deletions(-)
- 2 new files:
path-resolver.ts,path-resolver.test.ts
Full changelog: v5.9.0...v5.9.1
v5.9.0
Added
- Architecture & Design Support - Integrated architectural decision-making into core workflows:
/clavix:start: Added explicit probing for architectural patterns and design choices during conversation/clavix:summarize: Added "Architecture & Design" section to requirements extraction and Mini-PRD template/clavix:prd: Added dedicated Question 3.5 to structured Q&A for capturing design patterns (Monolith vs Microservices, Clean Architecture, etc.)/clavix:plan: Added "Architecture First" principle to task generation, prioritizing structural setup tasks
Changed
- Agentic Template Improvements:
- Simplified optional questions in
/clavix:prdby removing "Press Enter to skip" instructions (agents handle optionality natively) - Updated Plan template to enforce architectural setup before implementation tasks
- Simplified optional questions in
v5.8.2 - Clarifying Questions Protocol + Template Customization Guide
🎯 Highlights
Clarifying Questions Protocol - AI agents now systematically gather critical information with 95% confidence threshold
Template Customization Guide - Simple documentation for customizing Clavix commands
✨ Added
Clarifying Questions Protocol
- New systematic protocol for agents to ask clarifying questions
- 95% confidence threshold - only ask when genuinely needed
- Structured formats for multiple choice and custom input questions
- Integrated into AGENT_MANUAL for universal access across all commands
- Applied to planning/improvement commands:
improve,prd,plan,start,summarize - Best practices and examples (good vs bad questions)
- Recovery pattern for realizing clarification is needed mid-execution
Template Customization Documentation
- New
docs/template-customization.mdwith practical, straightforward guide - Complete reference table showing where each of 16 integrations generates templates
- Simple 3-step workflow: find file → edit → save
- Real-world customization examples (project context, security checklists, team conventions)
- No overcomplication - users edit generated templates directly in their integration directories
🔄 Changed
- MANIFEST.md - Updated component usage matrix with clarifying-questions component
- All canonical templates - Version references updated to v5.8.2
📊 Quality
- ✅ Build: Clean with all consistency checks passing
- ✅ Linter: Passed (0 errors)
- ✅ Tests: 1297 passing
🚀 Upgrade
npm install -g clavix@5.8.2
clavix updateNote: All changes are template/documentation-only. Zero TypeScript code changes, zero breaking changes.
🧹 Maintenance
- Removed obsolete legacy cleanup tests (feature was removed)
- Fixed lint warnings in integrations helper (no casts)
- Verified: 53 suites, 1297 tests passing; linter clean
v5.8.1: Technical Audit & Cleanup
Key Changes
- Legacy Cleanup: Removed deprecated v4 migration logic.
- Template Standardization: All templates now use shared components for transparency and protocols.
- Audit Mode:
/clavix:verifynow performs a spec-driven code audit. - Docs: Updated command reference and architecture.