Skip to content

CHANGELOG.md

Latest commit

 

History

History
764 lines (539 loc) · 32.6 KB

CHANGELOG.md

File metadata and controls

764 lines (539 loc) · 32.6 KB

Changelog

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

[7.2.1] - 2026-01-27

Fixed

  • Missing Skill Templates in Package - Added skill templates to npm package that were missing from v7.2.0:
    • using-clavix.md meta-skill template
    • implement-templates/ directory with subagent prompt templates
    • Updated implement.md, verify.md, plan.md, prd.md skill templates with verification patterns

[7.2.0] - 2026-01-27

Added

  • Verification Loop Patterns - Robust verification and fix loops across Clavix skills:

    • Iron Laws: "No completion without verification evidence" and "Issues found = issues fixed + re-verified"
    • Verification Gate Pattern: IDENTIFY → RUN → READ → VERIFY before any completion claims
    • Mandatory Fix Loops: Cannot proceed until issues are fixed AND re-verified
    • 3-Strike Rule: If 3+ fix attempts fail, stop and ask for help

  • using-clavix Meta-Skill - New orchestration skill establishing workflow rules:

    • Skill invocation requirements (check for skills BEFORE any action)
    • Complete workflow map showing skill chains
    • Required skill chains (prd → plan → implement → verify)
    • Red flags table for catching rationalization patterns
    • Skill priority order (exploration → planning → implementation → verification)

  • Subagent Templates - New templates for delegating implementation work:

    • clavix-implement/implementer-prompt.md - Dispatch implementer subagent
    • clavix-implement/spec-reviewer-prompt.md - Dispatch spec compliance reviewer
    • clavix-implement/quality-reviewer-prompt.md - Dispatch code quality reviewer
    • Two-stage review pattern: spec compliance first, then code quality

  • Required Sub-Skill References - Explicit skill chain enforcement:

    • clavix-prd → REQUIRED: clavix-plan
    • clavix-plan → REQUIRED: clavix-implement
    • clavix-implement → REQUIRED: clavix-verify (after all tasks)

Changed

  • clavix-implement Skill - Major verification enhancements:

    • Added Iron Law section at top of skill
    • Replaced optional verification with mandatory Verification Gate
    • Added Fix Loop pattern with re-verification requirement
    • Added subagent template references
    • Removed "skip verification" options from troubleshooting

  • clavix-verify Skill - Mandatory fix loops:

    • Added Iron Law: "Issues found = issues fixed + re-verified"
    • Changed "Fixing Workflow (Optional Loop)" to "Fixing Workflow (MANDATORY Loop)"
    • Added evidence requirement for claiming fixes
    • Added Red Flags table for catching skip temptations
    • Enhanced After Verification section with mandatory fix loop diagram

  • clavix-plan Skill - Skill chain enforcement:

    • Added REQUIRED SUB-SKILL header template for generated plans
    • Added skill chain diagram in Next Steps section

  • clavix-prd Skill - Skill chain enforcement:

    • Added REQUIRED SUB-SKILL reference to clavix-plan
    • Added complete skill chain diagram
    • Updated recommendations table to emphasize verification

Why v7.2.0?

This is a minor version bump because:

  1. Introduces new meta-skill (using-clavix) that orchestrates other skills
  2. Adds new subagent template files to clavix-implement
  3. Significantly enhances verification rigor across multiple skills
  4. Establishes mandatory skill chains (breaking previous optional patterns)

[7.1.1] - 2026-01-23

Fixed

  • TOML Parsing Error - Fixed escape character in prompt ID format ({std\\|comp} → {std\|comp}) that caused gemini-cli to fail parsing improve.toml (#16)

[7.1.0] - 2026-01-22

Added

  • Custom Agent Skills Path - New "Agent Skills - Custom Path" option in clavix init:
    • Users can specify any directory for skills installation (e.g., .aider-desk/skills)
    • Supports both relative paths (from project root) and absolute paths (including ~/)
    • Interactive prompt guides path type selection with clear explanations
    • Path stored in config.experimental.integrationPaths['agent-skills-custom']

[7.0.0] - 2026-01-21

BREAKING: Agent Skills as Primary Integration

Clavix v7 introduces Agent Skills as the recommended way to use Clavix. Agent Skills follow the agentskills.io specification and work with any compatible AI tool.

Why Agent Skills?

  • Universal compatibility - Works with any AI tool that supports the Agent Skills spec
  • Two installation scopes - Global (~/.config/agents/skills/) or Project (.skills/)
  • Efficient context - Skills optimized for < 500 lines each
  • Future-proof - As more tools adopt Agent Skills, your workflows work everywhere

Added

  • Agent Skills Integration - Full support for agentskills.io specification:

    • New AgentSkillsAdapter generates skills following the Agent Skills standard
    • Two installation scopes: Global (~/.config/agents/skills/) and Project (.skills/)
    • All 10 Clavix workflows available as curated skills
    • Skills are directory-based with SKILL.md files per agentskills.io spec
    • New "Agent Skills" section appears FIRST in integration selector during clavix init
    • Skills have optimized descriptions for agent discovery

  • Curated Skill Templates - New src/templates/skills/ directory:

    • 10 condensed, optimized skill templates based on canonical commands
    • Skills follow < 500 line recommendation for efficient context usage
    • Each skill includes proper YAML frontmatter (name, description, license)
    • Skills include: State Assertions, Self-Correction Protocols, Mode Boundaries, File-Saving Protocols, and Workflow Navigation

Fixed

  • Preserve existing command files during init - clavix init no longer overwrites user's existing command files in shared directories. Only files matching the clavix-* prefix pattern are updated.

Changed

  • README revamp - Skills-first documentation with cleaner, leaner structure
  • Integration selector priority - Agent Skills now appears first for better discoverability

[6.1.0] - 2026-01-12

Added

  • GitHub Copilot Custom Slash Commands - Full integration with VS Code GitHub Copilot:
    • Replaced global .github/copilot-instructions.md with proper slash commands
    • New adapter generates .github/prompts/*.prompt.md files
    • Commands appear as /clavix-improve, /clavix-prd, etc. in Copilot Chat
    • YAML frontmatter with name, description, agent, and tools
    • Smart agent mapping: ask for planning commands, agent for implementation
    • Moved from "Universal Adapters" to "IDE & IDE Extensions" category

Changed

  • CopilotPromptsAdapter replaces CopilotInstructionsGenerator
  • Integration name: copilot-instructions → copilot
  • File location: .github/copilot-instructions.md → .github/prompts/clavix-*.prompt.md
  • Schema updates: Added .prompt.md extension and prompt-files specialAdapter type

Removed

  • CopilotInstructionsGenerator class
  • src/templates/agents/copilot-instructions.md template

[6.0.0] - 2026-01-12

Added

  • New /clavix:review Command - Criteria-driven PR review for team collaboration (Inspired by @fmflurry):

    • Review teammate's PRs without requiring them to use Clavix
    • Conversational intake: asks which branch/PR, which criteria, and team conventions
    • 5 Review Presets: Security, Architecture, Standards, Performance, All-Around
    • Custom criteria support: describe specific concerns in natural language
    • Structured output: Executive Summary + Detailed Findings with severity levels
    • Severity categories: 🔴 Critical, 🟠 Major, 🟡 Minor, ⚪ Suggestion
    • Saves reports to .clavix/outputs/reviews/ with frontmatter metadata
    • Differentiates from /clavix:verify (which checks YOUR code against YOUR PRD)

  • New Template Components:

    • references/review-criteria.md - Comprehensive review dimension definitions
    • sections/review-presets.md - Predefined criteria preset configurations
    • sections/review-examples.md - Example review report outputs

  • Comprehensive Test Suite - 32 new tests for review command:

    • Template structure validation
    • Component existence and content checks
    • MANIFEST and documentation verification
    • Build artifact verification

Changed

  • Slash Commands: Now 10 total (was 9) with addition of /clavix:review
  • MANIFEST.md: Updated with review command and its components
  • cli-reference.md: Added reviews directory to file structure
  • docs/commands.md: Full documentation for review command with examples

Why v6.0.0?

This is a major version bump because:

  1. Introduces a significant new workflow (PR review) that expands Clavix beyond personal use to team collaboration
  2. Changes the command count from 9 to 10
  3. Adds new output directory structure (.clavix/outputs/reviews/)

[5.10.3] - 2026-01-09

Fixed

  • Cursor Integration Path - Commands now correctly write to .cursor/commands/ instead of .cursor/rules/

[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

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

[5.10.2] - 2026-01-05

Added

  • Task Selection for Implement Command - New task selection mode:
    • all - Execute all tasks sequentially
    • task N - Execute specific task by number
    • Interactive list - Select task from numbered list
    • Confirmation prompt before starting implementation
    • Better control over which tasks to execute

[5.10.1] - 2026-01-04

Fixed

  • Codex Prompts Subdirectory - Codex prompts now correctly use /prompts subdirectory with $CODEX_HOME environment variable

[5.9.2] - 2026-01-01

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

[5.9.1] - 2025-12-31

Added

  • $CODEX_HOME Environment Variable Support - Custom Codex prompts directory:
    • New src/utils/path-resolver.ts for centralized path resolution with priority system
    • Environment variable priority: $CODEX_HOME → user config → default path
    • Interactive prompting during clavix init when $CODEX_HOME is detected
    • Custom paths stored in .clavix/config.json under experimental.integrationPaths
    • Extensible pattern for future integrations (e.g., CURSOR_HOME, GEMINI_HOME)
    • 13 new comprehensive tests (all passing)

Changed

  • UniversalAdapter - Updated to use new path resolver instead of direct tilde expansion
  • AgentManager - Now accepts and passes userConfig to all adapters
  • All Adapters - Updated constructors to accept optional userConfig parameter:
    • ClaudeCodeAdapter
    • GeminiAdapter, QwenAdapter, LlxprtAdapter (via TomlFormattingAdapter)
    • UniversalAdapter

Technical Details

  • Added IntegrationPathsConfig interface to src/types/config.ts
  • Added schema validation for integrationPaths in src/utils/schemas.ts
  • Updated clavix init to detect and prompt about $CODEX_HOME
  • Updated clavix update to pass config to AgentManager
  • Backward compatible - existing installations work without changes

[5.9.0] - 2025-12-23

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:prd by removing "Press Enter to skip" instructions (agents handle optionality natively)
    • Updated Plan template to enforce architectural setup before implementation tasks

[5.8.2] - 2025-12-07

Added

  • Clarifying Questions Protocol - New systematic protocol for gathering critical information:

    • Added clarifying-questions.md component in agent-protocols/
    • Integrated into AGENT_MANUAL for universal access
    • Applied to planning/improvement commands: improve, prd, plan, start, summarize
    • 95% confidence threshold for when to ask questions
    • Structured format for multiple choice and custom input questions
    • Best practices and examples for effective questioning

  • Template Customization Documentation - Comprehensive guide for customizing Clavix templates:

    • New docs/template-customization.md with full integration paths
    • Complete reference table showing where each integration generates templates
    • Step-by-step workflows for customizing canonical templates, integration-specific overrides, and components
    • Common customization examples (project context, security checklists, team standards)
    • Troubleshooting guide for template customization issues

Changed

  • MANIFEST.md - Updated component usage matrix:
    • Added clarifying-questions.md to agent protocols section
    • Updated usage matrix to show which commands include the new protocol
    • Clarified that clarifying-questions is included via AGENT_MANUAL

[5.6.6] - 2025-11-28

Added

  • Test Coverage Improvements - 21 new tests for CLI commands:

    • 10 command behavior verification tests for update.ts
    • 11 reconfiguration flow tests for init.ts
    • Total test count: 1306 tests passing

  • Template Enhancements:

    • Task dependency guidance in plan.md with explicit dependency markers
    • Scope creep detection and handling in start.md
    • Archive size management guidance in archive.md
    • Multi-topic handling in refine.md troubleshooting

  • Quality Dimension Documentation:

    • Workflow-specific dimension usage table in quality-dimensions.md
    • Documented why /clavix:summarize excludes Specificity dimension
    • Added quality dimensions reference to refine.md

Fixed

  • Documentation Consistency:

    • Updated package.json description to list all 9 slash commands
    • Fixed "19+" to "20" AI tools count in README.md and docs
    • Standardized quality dimension terminology in refine.md (was using "Context/Constraints" instead of standard 6 dimensions)

  • MANIFEST.md - Added /clavix:refine to usage matrix and updated quality-dimensions description

[5.6.5] - 2025-11-28

Fixed

  • UserConfigSchema Missing Fields - Fixed Zod schema to match ClavixConfig interface

[5.6.4] - 2025-11-28

Fixed

  • UserConfigSchema Missing Fields - Fixed Zod schema to match ClavixConfig interface:
    • Added version, templates, outputs, preferences, experimental fields
    • Eliminates false "Unknown fields" warnings during clavix update

[5.6.3] - 2025-11-28

Added

  • Zod Schema Validation - Runtime validation for configuration files:

    • Added zod dependency for robust schema validation
    • Created src/utils/schemas.ts with validation schemas for integrations.json and user config.json
    • Integrated validation at adapter registry load time and in CLI commands
    • Contextual error messages for configuration issues

  • Universal Adapters in integrations.json - 4 new documentation-only integrations:

    • agents-md - AGENTS.md universal agent guidance
    • copilot-instructions - GitHub Copilot instructions
    • octo-md - OCTO.md octopus format
    • warp-md - WARP.md format
    • Added type field ("standard" or "universal") to all integration entries

  • Template Documentation - New docs/templates.md authoring guide:

    • Component include system ({{INCLUDE:}} markers)
    • Template anatomy and structure
    • Quality checklist for contributors

  • Schema Validation Tests - 23 new tests covering:

    • IntegrationEntrySchema, IntegrationsConfigSchema, UserConfigSchema
    • Validation functions and error formatting

Changed

  • Summarize→Plan Workflow - Improved template coordination:

    • Added "Suggest + Confirm" pattern for project name in summarize template
    • Added fallback detection for summarize/ directory in plan template
    • Added confirmation message with task count in implement template

  • Error Output Standardization - Replaced console.log/console.error with Oclif methods:

    • All CLI commands now use this.log() and this.error() for consistent output
    • Contextual error hints for common configuration issues

  • Template Header Standardization - All 9 slash command templates now use consistent format:

    • "## State Assertion (REQUIRED)" section header
    • Consistent structure across improve, prd, plan, implement, start, summarize, refine, verify, archive

  • README.md Reorganization - Workflow-oriented documentation:

    • Added Mermaid workflow diagram showing command relationships
    • Organized commands by workflow: Quick Path, Full Planning, Exploratory, Refinement
    • All 9 slash commands prominently listed

  • CONTRIBUTING.md Updates - Enhanced contributor guide:

    • Added "Quick Start for First-Time Contributors" decision table
    • Emphasized template-first architecture rules

Fixed

  • JSON Parsing Crashes - Wrapped all JSON.parse calls with try-catch:

    • update.ts, init.ts, version.ts now handle malformed JSON gracefully
    • User-friendly error messages instead of stack traces

  • Test Suite Updates - Fixed tests for adapter count changes (16 → 20):

    • agent-manager.test.ts - Updated expected adapter count
    • adapter-interface.test.ts - Added special case for root directory adapters
    • multi-integration-workflow.test.ts - Added detection markers for universal adapters
    • init.test.ts - Fixed error output assertion

[5.5.2] - 2025-11-28

Added

  • /clavix:refine Command - New slash command for refining existing PRDs and prompts:

    • Auto-detects available refinement targets (PRDs and saved prompts)
    • Change tracking with [ADDED], [MODIFIED], [REMOVED], [UNCHANGED] markers
    • Before/after quality comparison for prompts
    • Refinement history tracking in PRD files
    • Integration guidance for next steps (plan regeneration, implementation)

  • Mandatory AGENTS.md Integration - Universal agent guidance is now always enabled:

    • AGENTS.md automatically included in all installations
    • Removed from user selection (handled internally)
    • Added ensureMandatoryIntegrations() helper function
    • Users informed during init: "AGENTS.md is always enabled to provide universal agent guidance"

  • New AgentErrorMessages Methods - Extended error messaging for better diagnostics:

    • templateNotFound() - When template files are missing
    • adapterNotFound() - When adapter lookup fails
    • configLoadFailed() - When configuration parsing fails
    • updateFailed() - When update command encounters issues
    • diagnosticFailed() - When diagnose command finds problems

Changed

  • Error Handling Consistency - Replaced generic Error throws with typed DataError:

    • agents-md-generator.ts, octo-md-generator.ts, warp-md-generator.ts
    • copilot-instructions-generator.ts, instructions-generator.ts
    • toml-templates.ts

  • Constants Consolidation - Cleaned up src/constants.ts:

    • Removed 11 unused constants
    • Kept only CLAVIX_BLOCK_START and CLAVIX_BLOCK_END
    • Updated all imports to use constants instead of hardcoded values

  • CONTRIBUTING.md Updates - Added explicit architecture boundaries:

    • "Explicitly Forbidden Features" section with 4 rejected proposals
    • Clear explanations for WHY certain features don't fit the architecture
    • Guidance on what TO do instead of forbidden patterns

Fixed

  • Documentation Accuracy - Fixed coverage claims and output paths:
    • Updated TESTING.md: Changed 100% to 70%+ (actual thresholds)
    • Updated getting-started.md: Added summarize outputs documentation
    • Updated docs/commands.md with full refine command documentation

[5.5.1] - 2025-11-28

Added

  • CONTRIBUTING.md - Comprehensive contributor guide with architecture principles:

    • Documents agentic-first architecture and why it must remain so
    • Explains what CAN and CANNOT be changed (TypeScript won't help slash commands)
    • Full development setup, testing instructions, and PR process

  • Command Format Visibility - Users now see their command format prominently:

    • clavix init output shows format at the TOP (e.g., "Your command format: /clavix:improve")
    • Generated INSTRUCTIONS.md includes command format table
    • All docs include format reference tables

Changed

  • Documentation Consolidation - Reduced 18+ scattered files to 5 focused documents:

    • docs/README.md - Single entry point with navigation
    • docs/architecture.md - Consolidated from how-it-works.md + philosophy.md
    • docs/commands.md - All 12 command docs in one file
    • docs/getting-started.md - Consolidated from 4 guide files
    • docs/integrations.md - Added Format column to integration tables

  • README.md - Slimmed down with prominent command format section and links to consolidated docs

Removed

  • Deleted docs/commands/ directory (12 individual command files)
  • Deleted docs/guides/ directory (4 guide files)
  • Deleted docs/how-it-works.md, docs/philosophy.md, docs/why-clavix.md (merged into architecture.md)

[5.5.0] - 2025-11-27

Changed

  • Adapter Architecture Refactor - Unified adapter system using config-driven factory pattern:

    • Migrated 12 simple adapters to UniversalAdapter class with ADAPTER_CONFIGS registry
    • Kept dedicated classes for special adapters: ClaudeCodeAdapter (doc injection), GeminiAdapter, QwenAdapter, LlxprtAdapter (TOML format)
    • Reduced code duplication significantly while maintaining all functionality

  • Error Handling Improvements:

    • Template assembly failures now throw DataError instead of silent warnings
    • Standardized all console.warn calls to use logger.warn utility
    • Fixed deprecated escapeRegex wrapper in base-adapter.ts

Removed

  • Code Consolidation:

    • Deleted 12 redundant adapter source files (cursor, windsurf, kilocode, roocode, cline, droid, opencode, crush, codebuddy, amp, augment, codex)
    • Deleted corresponding 12 adapter test files
    • Removed deprecated archive documentation files

  • Documentation Cleanup:

    • Removed v4-era configuration references from docs/guides/configuration.md
    • Removed outdated session references from command documentation
    • Deleted docs/archive/ directory with obsolete files

Fixed

  • Updated test files to use dynamic adapter paths instead of hardcoded directories
  • Fixed adapter contract tests to work with new factory pattern
  • Fixed multi-integration workflow tests for correct adapter detection

[5.4.0] - 2025-11-27

Changed

  • Documentation cleanup - Comprehensive review and update of all documentation:

    • Removed outdated /docs/clavix-intelligence.md (referenced removed pattern system)
    • Updated /docs/guides/workflows.md for v5 architecture (no more clavix fast/clavix deep)
    • Updated /docs/how-it-works.md to explain agentic-first architecture
    • Fixed /src/templates/instructions/README.md component references

  • Consistent terminology - Standardized naming across all files:

    • "Optimize" → "Improve" in template naming and documentation
    • "provider" → "integration" in code comments and variables
    • "execute" → "implement" in agent templates

  • Template improvements - Enhanced agent instruction quality:

    • Added Self-Correction Protocol to implement.md, verify.md, archive.md
    • Added State Assertion blocks for mode enforcement
    • Softened defensive tone in improve.md while maintaining mode boundaries

Removed

  • Dead code cleanup:

    • Removed preserveSessions from PreferencesConfig (sessions removed in v5.3.0)
    • Removed sessionNotFound() from AgentErrorMessages
    • Removed session directory references from error messages

  • Outdated documentation:

    • Deleted /docs/clavix-intelligence.md
    • Removed broken links from /docs/README.md

Fixed

  • Test factory alignment - Updated config-factory.ts to match current ClavixConfig shape
  • Instructions README - Fixed references to non-existent component files

[5.3.1] - 2025-11-27

Fixed

  • Cline integration cleanup bug - When deselecting Cline during clavix init reconfiguration and choosing "Clean up", the .clinerules/workflows/clavix/ subdirectory is now properly removed. Previously, removeAllCommands() only removed files, ignoring subdirectories.

[5.3.0] - 2025-11-27

BREAKING CHANGES

  • Session persistence removed - The .clavix/sessions/ directory is no longer created or used. Session-based workflows are now handled entirely by AI agent context.

Added

  • Adapter registry infrastructure - New config-driven adapter system (adapter-registry.ts, universal-adapter.ts) enables consistent adapter behavior through configuration rather than code duplication.

  • Init reconfiguration menu - When running clavix init in an already-initialized project, a menu offers three options:

    • "Reconfigure integrations" - Change selected integrations
    • "Update existing" - Regenerate commands for current integrations
    • "Cancel" - Exit without changes

  • Removed commands consistency test - New test suite (removed-commands.test.ts) verifies deprecated commands are properly removed from CLI, manifest, and documentation.

  • Template component manifest - New MANIFEST.md documents all reusable components in the template system.

Changed

  • Config command eliminated - clavix config functionality has been merged into clavix init:

    • Use clavix init → "Reconfigure integrations" to change integrations
    • Use clavix init → "Update existing" to regenerate commands

  • Legacy cleanup utility deprecated - legacy-command-cleanup.ts is now marked for removal in v6.0.0 as the transition period for old naming patterns completes.

Removed

  • clavix config command - Use clavix init instead (see Changed section)

  • Orphaned template components - Removed unused decision-rules.md and error-handling.md from _components/

  • Session references in templates - Removed session-related content from agents.md, copilot-instructions.md, warp.md, plan.md, and file-operations.md

Fixed

  • Test suite improvements - Added comprehensive tests for:
    • Update command flag behavior (docs-only, commands-only, force)
    • Init reconfiguration menu flows
    • v5.3 removed features (sessions directory, config command)

[5.2.1] - 2025-11-27

Fixed

  • Diagnose command false warnings - Fixed 3 bugs where clavix diagnose showed incorrect warnings after successful clavix init:
    • No longer expects .clavix/commands/ directory (commands go to adapter-specific directories)
    • Recognizes doc generator integrations (agents-md, octo-md, warp-md, copilot-instructions) as valid
    • Removed misleading "No slash commands installed" check

[5.2.0] - 2025-11-27

Added

  • New clavix diagnose command - Full diagnostic report for troubleshooting installations

    • Version check
    • Directory structure validation
    • Config integrity verification
    • Integration status with command counts
    • Template integrity check
    • Summary with recommendations

  • Slash commands in --help - Help output now shows available slash commands alongside CLI commands

  • Feature matrix in README - Clear comparison of capabilities across integrations (Claude Code, Cursor, Gemini, etc.)

Changed

  • DRY adapter architecture - Created TomlFormattingAdapter base class eliminating ~140 lines of duplication across Gemini, Qwen, and LLXPRT adapters

  • DRY documentation generators - Refactored AgentsMdGenerator, OctoMdGenerator, WarpMdGenerator, and CopilotInstructionsGenerator to use DocInjector utility

  • Verify/Archive reorganized - Now clearly documented as "Agentic Utilities" separate from core workflow commands

Removed

  • Global config flag - Removed unimplemented -g/--global flag from clavix config command

  • Vestigial v4 config types - Removed ~90 lines of unused IntelligenceConfig, EscalationThresholdsConfig, QualityWeightsConfig, and PatternSettingsConfig interfaces

Fixed

  • Consistent documentation - All agent templates (CLAUDE.md, AGENTS.md, OCTO.md, WARP.md, copilot-instructions.md) now have consistent verify/archive categorization

[5.1.1] - 2025-11-27

Changed

  • Consolidated execute into implement command
  • Removed standalone prompts command (now part of implement workflow)

[5.0.0] - 2025-11-27

BREAKING: Agentic-First Architecture

Clavix v5 is a complete architectural rewrite. The TypeScript intelligence layer has been removed in favor of a lean, agentic-first design where markdown templates are the product.

Why This Change?

Analysis revealed that 99% of Clavix usage was through slash commands in AI IDEs (Claude Code, Cursor, etc.), where AI agents read markdown templates and execute them using their native tools. The TypeScript intelligence layer (~18,500 lines) only executed via rarely-used CLI commands.

v5 embraces reality: templates instruct agents directly, no TypeScript code runs during workflow execution.

What's Removed

Component Lines Removed
CLI commands (15) ~5,800
Intelligence layer ~7,790
Core managers (14) ~5,200
Tests for removed code ~31,300
Total ~50,000 lines
Removed CLI Commands
  • clavix improve, clavix prd, clavix plan, clavix implement
  • clavix start, clavix summarize, clavix execute, clavix verify
  • clavix archive, clavix analyze, clavix task-complete
  • clavix list, clavix show, clavix prompts list/clear
Removed Core Modules
  • universal-optimizer.ts, quality-assessor.ts, pattern-library.ts
  • intent-detector.ts, confidence-calculator.ts
  • 20 pattern files in intelligence/patterns/
  • prompt-manager.ts, session-manager.ts, task-manager.ts
  • archive-manager.ts, git-manager.ts, prd-generator.ts
  • verification-manager.ts, question-engine.ts

What Remains

4 CLI Commands (for setup only):

Command Purpose
clavix init Initialize Clavix in a project
clavix update Update templates after package update
clavix config Manage configuration
clavix version Show version

8 Slash Commands (executed by AI agents):

  • /clavix:improve - Prompt optimization
  • /clavix:prd - PRD generation
  • /clavix:plan - Task breakdown
  • /clavix:implement - Task/prompt execution (auto-detects source)
  • /clavix:start - Conversational exploration
  • /clavix:summarize - Extract requirements from conversation
  • /clavix:verify - Verify implementation
  • /clavix:archive - Archive completed projects

22 Integration Adapters - All adapters remain for multi-tool support.

How It Works Now

  1. User runs /clavix:improve "prompt" in their AI IDE
  2. AI agent reads the markdown template at .clavix/commands/improve.md
  3. Agent follows instructions using native tools (Write, Read, Edit, etc.)
  4. Output saved to .clavix/outputs/

No TypeScript code executes. Templates are self-contained instructions.

Template Updates

Templates updated for v5 agentic-first approach:

  • Removed CLI command references (agents use native tools)
  • Removed .index.json - prompts use frontmatter metadata
  • Added explicit "How I Do It" sections for agent operations
  • Updated file format documentation

Migration from v4

If you have v4 projects:

  1. Run npm install -g clavix@latest
  2. Run clavix update in your project
  3. Old outputs in .clavix/outputs/ are preserved
  4. v4 CLI commands no longer exist - use slash commands instead

For v4 Documentation

See docs/archive/v4-architecture.md for the previous architecture.