v7.3.0 - Clean Command

Full Changelog: v7.2.2...v7.3.0

v7.2.3

Fix: Ensure /clavix:improve Writes Prompt File

When running /clavix:improve, the improved prompt file was not being created at .clavix/outputs/prompts/{id}.md. This caused /clavix:implement --latest to fail because it could not find the latest prompt file.

Root Cause

The clavix-improve skill template had a descriptive rather than prescriptive "File-Saving Protocol". It showed the file format but didn't explicitly command the AI to invoke the Write tool.

Changes

• Rewrite File-Saving Protocol with explicit tool invocation commands
• Replace descriptive "Save File" with "You MUST use the Write tool"
• Remove bash mkdir example (Write tool creates parent dirs automatically)
• Add blocking checkpoint requiring verification before final output
• Emphasize that displayed content ≠ saved file

This matches the slash command format and ensures /clavix:implement --latest can find the latest prompt file.

v7.2.2

What's Changed

Superpowers Pattern for using-clavix Meta-Skill

The using-clavix skill now follows the superpowers pattern for proper workflow orchestration:

• Renamed to using-clavix (dropped the clavix- prefix) - this is the meta-skill that tells agents how to use all other Clavix skills
• AGENTS.md now instructs agents to load using-clavix FIRST before any Clavix workflow
• This enables the "1% rule" pattern: if there's even a 1% chance a Clavix skill applies, agents must check it first

How It Works

1. Agent reads AGENTS.md → sees "FIRST: Load using-clavix Skill"
2. Agent invokes using-clavix skill
3. Skill establishes:
   • Skill invocation rules (check skills BEFORE any response)
   • Required skill chains (prd → plan → implement → verify)
   • Iron Laws for verification (no completion claims without evidence)
   • Workflow orchestration and fix loops

Full Changelog

• feat: AgentSkillsAdapter treats using-clavix as special case (no prefix)
• feat: Updated AGENTS.md template with "FIRST: Load using-clavix" section
• test: Added test for using-clavix special naming

v7.2.1 - Skill Templates Fix

What's Fixed

Missing Skill Templates - v7.2.0 updated the package version and changelog but the actual skill template files were not included. This patch adds:

• using-clavix.md - Meta-skill for workflow orchestration
• implement-templates/ - Subagent prompt templates:
  • implementer-prompt.md
  • spec-reviewer-prompt.md
  • quality-reviewer-prompt.md
• Updated skill templates with verification patterns:
  • implement.md - Iron Law, Verification Gate, Fix Loop
  • verify.md - Mandatory fix loops, evidence requirements
  • plan.md - Skill chain references
  • prd.md - Skill chain references

Full Changelog: v7.2.0...v7.2.1

v7.2.0 - Verification Loop Patterns

What's New

Verification Loop Patterns

This release adds robust verification and fix loop patterns across Clavix skills, inspired by proven agentic workflow patterns.

Iron Laws

• No completion without verification evidence - If you haven't run the verification command, you cannot claim it passes
• Issues found = issues fixed + re-verified - Cannot proceed until fixed AND 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.

New using-clavix Meta-Skill

A new orchestration skill establishing:

• 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:

• implementer-prompt.md - Dispatch implementer subagent
• spec-reviewer-prompt.md - Dispatch spec compliance reviewer
• 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)

---

Full Changelog: v7.1.1...v7.2.0

v7.1.0 - Custom Agent Skills Path

What's New

Custom Agent Skills Path

New "Agent Skills - Custom Path" option in clavix init:

• 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']

Installation

npm install -g clavix@7.1.0
clavix init

Then select "Agent Skills - Custom Path" and provide your preferred directory.

Full Changelog: v7.0.0...v7.1.0

v7.0.0 - Agent Skills Integration

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 and Project
  • All 10 Clavix workflows available as curated skills
  • Skills are directory-based with SKILL.md files per spec
  • Agent Skills section appears FIRST in integration selector

• Curated Skill Templates - 10 condensed, optimized skill templates

  • < 500 line recommendation for efficient context usage
  • Proper YAML frontmatter (name, description, license)
  • State Assertions, Self-Correction Protocols, Mode Boundaries

Fixed

• Preserve existing command files during init - Only files matching clavix-* prefix are updated

Changed

• README revamp - Skills-first documentation with cleaner, leaner structure

---

Full Changelog: https://github.com/ClavixDev/Clavix/blob/main/CHANGELOG.md

v6.1.2

Bug Fixes

• Fix escape character in prompt ID format in improve command template (#16)

The pipe character in the prompt ID format {std|comp} was not properly escaped in markdown tables, causing it to render incorrectly. This fix ensures the backslash is preserved: {std\\|comp}.

Full Changelog: v6.1.1...v6.1.2

v6.1.1

What's Changed

• docs: Reference AskUserQuestion tool in AGENT_MANUAL - Explicitly mentions the AskUserQuestion tool when agents need clarification, replacing generic guidance

Full changelog: v6.1.0...v6.1.1

v6.1.0 - GitHub Copilot Custom Slash Commands

🚀 Highlights

GitHub Copilot now gets proper slash commands instead of global instructions! Commands appear as /clavix-improve, /clavix-prd, etc. directly in VS Code Copilot Chat.

✨ What's New

• Replaced .github/copilot-instructions.md with .github/prompts/*.prompt.md files
• Native slash commands in VS Code: /clavix-improve, /clavix-prd, /clavix-implement, etc.
• Smart agent mapping: Planning commands use ask agent, implementation uses agent with tools
• YAML frontmatter with name, description, agent, and tools configuration

⚠️ Breaking Change

• Integration renamed from copilot-instructions → copilot
• If you have copilot-instructions in your .clavix/config.json, update it to copilot
• Run clavix init to regenerate with the new format

📦 Installation

npm install -g clavix@latest
clavix init

Full Changelog: v6.0.0...v6.1.0