v2.0.2 - Fix skills-management trigger and post-search review

Changes

• Fix skills-management trigger: Updated description with explicit trigger phrases ("find a skill for X", "search for a skill", "is there a skill for X") so the skill is invoked automatically for discovery requests
• Fix post-search review behavior: Changed from only reviewing ambiguous results to always suggesting review of found skills after a search
• Bump plugin.json version to 2.0.2

v2.0.1 - Bug fixes & skill management improvements

Fixes

• Symlink crash fixed: shutil.rmtree failed on skills installed via npx skills add (symlinks). All 5 scripts now handle both symlinks and real directories correctly.

New features

• Multi-agent deletion: delete_skill.py now supports --agent (repeatable) and --all-agents flags. Default mode warns about remaining copies in other agent directories.
• Ecosystem skill management: Added npx skills remove/rm, npx skills list/ls, and step-by-step update workflow to SKILL.md.
• Always-update-all-agents guidance: SKILL.md now reminds the agent to push skill updates to all detected agents, not just Claude Code.

Housekeeping

• Renamed plugin directory from claude-code-reflection-skills to agents-reflection-skills
• Shortened skills-management YAML description (596 → 248 chars)

v2.0.0 - Universal Agent Support

What's New

Universal agent support

• Renamed all skills: dropped claude- prefix, manager → management
• Preparing for full cross-agent compatibility beyond Claude Code

MCP management across 10+ agents

• Integrated npx add-mcp for installing MCP servers to Claude Code, Cursor, VS Code, Claude Desktop, Gemini CLI, Codex, Goose, GitHub Copilot CLI, OpenCode, and Zed
• Handles config format differences (JSON, YAML, TOML) automatically
• New references/multi-agent.md with all agent config paths and field mappings

Codex CLI support

• settings-management: TOML config, approval policies, sandbox modes, profiles, Starlark rules
• hooks-management: notify setting, AfterAgent/AfterToolUse events, Starlark rules as PreToolUse alternative
• subagents-management: AGENTS.md system, skills, override files

42-agent skills registry

• skills-management now supports managing skills across 42 AI coding agents
• Detect installed agents, copy/move skills between them

Remote skill assessment

• New framework for evaluating ecosystem skills from skills.sh before installation
• Quality signals: install count, agent distribution, age, description, relevance
• Red flag detection: inflated installs, ads, unnecessary dependencies
• Agent-driven assessment with Recommended / Consider / Skip verdicts

Standardized references

• Claude-specific files prefixed with claude-*
• Codex-specific files prefixed with codex-*

v1.1.1 - Hooks manager improvements

Changes

claude-hooks-manager

• Added guidance to prefer script files for complex hooks instead of inline escaping
• New section: "Simple vs Complex Hooks" with complexity table
• Bash script template for hook development
• Recommended script location: ~/.claude/hooks/

Cleanup

• Removed redundant top-level /skills/ directory (all skills are in plugin)

v1.1.0 - Multi-Agent Support

What's New

Multi-Agent Skills Support

The skills-manager skill now supports 15 AI coding agents, allowing you to install, copy, and move skills between different AI tools.

Supported Agents:
OpenCode, Claude Code, Codex, Cursor, Amp, Kilo Code, Roo Code, Goose, Gemini CLI, Antigravity, GitHub Copilot, Clawdbot, Droid, Windsurf

New Commands

• detect_agents.py - Detect installed AI agents
• list_agent_skills.py - List skills for any agent
• install_skill.py - Install skills to one or more agents
• copy_skill.py - Copy skills between agents
• move_skill_agent.py - Move skills between agents

Breaking Changes

• Renamed claude-skills-manager to skills-manager

Acknowledgments

Multi-agent support is based on add-skill by Vercel Labs.

v1.0.1 - Improve claude-mcp-manager skill

Changes

• claude-mcp-manager skill improvements following best practices:
  • Reduced SKILL.md from 360 to 100 lines (~72% reduction)
  • Added table of contents to all reference files
  • Moved detailed configuration examples to scopes.md
  • Removed time-sensitive GitHub issue references
  • Consistent terminology throughout all files

Best Practices Applied

• Concise main SKILL.md that defers to reference files
• Progressive disclosure pattern for detailed content
• One-level deep references from SKILL.md
• No time-sensitive information

v1.0.0

Claude Code Reflection Skills v1.0.0

A marketplace for meta-skills that let Claude Code configure itself through conversation. No more editing JSON files.

Installation

# Add the marketplace
/plugin marketplace add https://github.com/CodeAlive-AI/claude-code-reflection-skills.git

# Install the plugin
/plugin install claude-code-reflection-skills@claude-code-reflection-skills

Use Cases

1. Connect to your database — Install PostgreSQL MCP server
2. Auto-format code — Run Prettier/Black after every edit
3. Auto-run tests — Run pytest when Python files change
4. Enable sandbox mode — 84% fewer permission prompts
5. Block access to secrets — Protect .env and sensitive files
6. Create a code reviewer — Read-only subagent with Opus
7. Switch to Opus — Better reasoning for complex tasks
8. Share team config — Project-scoped settings for everyone
9. Log all commands — Full audit trail
10. GitHub integration — Create PRs and manage issues
11. Update MCP credentials — Rotate API keys without reinstalling

Included Skills

• claude-mcp-installer — Connect to databases, GitHub, APIs
• claude-hooks-manager — Auto-format, auto-test, logging
• claude-settings-manager — Permissions, sandbox, model selection
• claude-subagents-manager — Specialized agents for specific tasks
• claude-skills-manager — Organize and share skills
• claude-plugins-manager — Package and publish plugins

See README for details.