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.
- Flat Config Structure - Moved
registriesandskillsto top-level fields in project, global, and template configs. Previouslyregistrieswas nested underskills(skills.registries); now bothregistriesandskillssit at the same level for consistency across all config contexts. - Template Registries Support - Init templates can now declare custom
registriesthat get saved to the project config duringai-devkit init --template. - Shared Registry Filter - Extracted duplicate registry-filtering logic from
ConfigManagerandGlobalConfigManagerinto a sharedfilterStringRecord()helper. - Removed Dead Code - Removed unused
getInstalledSkills()method,normalizeSkillsConfig()method,SkillsConfiginterface, andSkillRegistriesConfiginterface.
- The
skillsfield in.ai-devkit.jsonis now always a plain array of{ registry, name }objects. The previous object format ({ registries: {...}, installed: [...] }) is no longer supported. - The global config (
~/.ai-devkit/.ai-devkit.json) now uses a top-levelregistriesfield instead ofskills.registries.
- Skill Remove Config Cleanup - Fixed
skill removenot removing the skill entry from.ai-devkit.json. - Install Update Guard - Fixed install service updating config when there are no successful installs.
- Install Skills Update Guard - Fixed install service including skills in the update object when there are no successful skills.
- Install Skills Object Format - Fixed
installfailing when theskillsfield in config is an object with aninstalledarray rather than a plain array.
- MCP Server Config Standardization - Added
mcpServerssupport toinitandinstallflows with environment-specific generators (ClaudeCodeMcpGenerator,CodexMcpGenerator) for standardized MCP server config across AI agents (#61). - Channel Debug Flag - Added
--debugflag to thechannelcommand for verbose debugging output.
- Memory MCP Tool Names - Renamed MCP tool names from dot-separated (
memory.storeKnowledge) to underscore-separated (memory_storeKnowledge,memory_updateKnowledge,memory_searchKnowledge) to comply with strict MCP client naming regex^[a-zA-Z0-9_-]{1,64}$; backward-compat aliases retained for existing users (#59). - Skill Add With No Environments - Fixed
skill addcrashing when no environments are defined in config. - Config addSkill Crash - Fixed
addSkill()crashing when theskillsfield is an object containing aregistrieskey rather than an array.
- Channel Connector Package - Added new
@ai-devkit/channel-connectorpackage with channel management, config storage, and Telegram adapter support, plus CLI commands for channel operations (#53). - Channel Connector Release Workflow - Added publish workflow and package ignore rules for releasing the channel connector package.
- Codex Plugin Config - Added
.codex-plugin/plugin.jsonconfiguration. - Dev Lifecycle Guide - Added web documentation for the dev-lifecycle skill.
- Shopify Toolkit Registry Entry - Added the Shopify toolkit skill to the registry.
- Git Hooks - Added Husky
pre-commitandpre-pushhooks.
- Memory Skill Workflow - Improved the memory skill workflow and updated the OpenAI agent configuration used by the skill.
- Dev Lifecycle Brainstorming - Updated the dev-lifecycle
new-requirementguidance to include brainstorming refinements. - Code Review Guidance - Refreshed code review instructions across command templates and dev-lifecycle references.
- CLI Environment Support - Added Amp Code environment configuration support and updated related CLI expectations.
- Interactive Skill Add Parsing - Fixed
ai-devkit skill add <registry>so it no longer requiresskill-nameand can open the interactive skill selection flow.
- Interactive Skill Selection -
ai-devkit skill addcan now present an interactive multi-select flow when adding skills from a registry (#51). - Memory DB Path Configuration - Added support for configuring a project-specific memory database path and using it in CLI and memory API flows (#50).
- Senior Engineer Skills - Added
verifyandtddskills and included them in thesenior-engineertemplate. - Targeted Global Skill Install Prompt - Added environment-aware prompting to support targeted global skill installation.
- E2E Coverage - Added end-to-end test coverage for the new memory database path configuration flow.
- Skill Guidance - Updated memory skill instructions and refreshed skill red-flag guidance.
- Bundled Skills Data - Added new bundled skills and updated skill registry metadata.
- E2E Test Maintenance - Cleaned up the end-to-end test suite for the new CLI flows.
- Ora Compatibility - Fixed the
oradependency/version issue.
- Agent Orchestration Skill - New
agent-orchestrationskill for coordinating multi-agent workflows, including OpenAI agent configuration.
- TTY Writer - Fixed sending enter separately for more reliable agent input delivery.
- Dependency Updates - Upgraded project dependencies.
- Analytics Config - Disabled Nx analytics in project configuration.
- Agent Detail Command - New
ai-devkit agent detail --id <name>command to inspect running agent conversations (#49). - Skill Registry - Added
samber/cc-skills-golangskill repository.
- Agent Identifier - Updated agent identifier; removed
slugfield fromAgentInfoin favor of simplified name-based matching.
- Agent List CWD - Agent list command now displays the current working directory for each running agent (#47).
- Clarification & Brainstorming Loop - Added clarification and brainstorming loop to
review-designandreview-requirementscommands.
- Generalized Session Mapping - Refactored process-to-session mapping into shared utilities (
matching,session,process) used by both Claude Code and Codex adapters, replacing adapter-specific implementations (#45). - Claude Sessions PID Matching - Updated Claude Code session matching for more reliable PID-based detection (#48).
- Custom Docs Directory - Added support for configuring a custom AI documentation directory via
paths.docsandai-devkit init --docs-dir <path>.
- Claude Code Adapter - Reimplemented Claude Code session matching to use process start times, bounded session scanning, and direct session-based summaries for more reliable agent detection.
- Node.js Requirement - Updated the minimum supported Node.js version to
20.20.0.
- Codex Cross-Repo Matching - Prevented Codex agent session matching from incorrectly attaching sessions across repositories.
- Git Passphrase Prompts - Fixed skill registry git operations so passphrase input works correctly during clone and update flows.
- Agent Send Command - New
ai-devkit agent sendcommand for sending input to running agents via TTY writer. - Agent Type Display - Agent list now shows agent type (e.g., Claude Code, Codex) in the listing output.
- Worktrees Location - Updated dev-lifecycle skill worktree setup references.
- Codex Adapter - Added Codex adapter support.
- Agent Manager Package - Added standalone
@ai-devkit/agent-managerpackage.
- CLI Agent Migration - Migrated
agentcommand to@ai-devkit/agent-manager. - Skill Registry Priority - Updated skill registry priority handling.
- Init/Install Templates - Stopped copying context templates during
init/install. - Version Output Source -
--versionnow uses package version output. - Documentation - Updated docs and improved web agent setup guide with template-based init.
- Project Templates - Added and updated templates used across setup flows.
- Install Validation - Added install smoke-test updates for
ai-devkit install.
- Agent List Display - Kept the
working-oncolumn to one line in agent list output. - Claude PID Session Mapping - Prefer exact history cwd for Claude pid-session mapping.
- Config Manager Phase Handling - Guarded missing phases in config manager.
- Docs Typos/Troubleshooting - Fixed typo and added Codex sandbox
npxtroubleshooting FAQ.
- Install Command - Added
ai-devkit installto apply project configuration from.ai-devkit.json- Supports
--config <path>for custom config file locations - Supports
--overwritefor non-interactive full overwrite mode - Installs environments, phases, and skills in a single run with summary output
- Supports
- Dependency Updates - Upgraded
better-sqlite3
- Lint Command - Added
ai-devkit lintcommand support - Template Mode for Init - Added init template mode with YAML/JSON support
- Memory Update Command - Added
ai-devkit memory updatefor modifying knowledge items by ID - New Skills - Added
capture-knowledge,simplify-implementationandtechnical-writerskills - Plugin Support - Added
.claude-pluginand.cursor-pluginintegration files
- Dev Lifecycle Workflows - Refactored worktree setup and new-requirement flow
- Lifecycle Documentation - Updated docs to require feature worktrees and make bootstrap language-agnostic
- Web Docs Navigation - Added linkable anchors for documentation section headings
- Command Templates - Updated CLI command templates
- Skill Registry Handling - Refresh cached skill registry automatically on
skill add - Documentation Updates - Refreshed README and development docs for current CLI behavior
- Memory Test Stability - Fixed flaky
updated_attimestamp test in memory module
- Dev Lifecycle Skill - Added structured SDLC skill with phase references and helper scripts
- Debug Skill - Added reusable debug skill definitions for agent workflows
- Web Skill Search Experience - Added
/skillsweb page and related docs/navigation updates - Memory Search Table Output - Added
ai-devkit memory search --tablefor terminal-friendly results
- Skill Registry Data - Updated skill registry/index content and automated rebuild outputs
- Documentation - Added/updated AI phase docs for setup wizard, web skill search, and memory search table output
- Init Environment Parsing - Improved
init -ehandling for full environment values
- Skill Search - New
skill findcommand to discover skills across all registries- Keyword Search: Find skills by name or description (e.g.,
ai-devkit skill find typescript)
- Keyword Search: Find skills by name or description (e.g.,
- Skill Index Rebuild - New
skill rebuild-indexcommand for search feature
- Native Fetch - Migrated network calls from
httpsto nativefetchAPI for cleaner code - GITHUB_TOKEN Support - GitHub API calls now use
GITHUB_TOKENenvironment variable when available
- Agent Management - Detect and control external AI agents
- List Agents:
ai-devkit agent list- View running agents (Claude Code, etc.) - Open Agent:
ai-devkit agent open <name>- Focus agent terminal window - Terminal Support: Works with tmux, iTerm2, and Apple Terminal
- Fuzzy Matching: Open agents by partial name
- List Agents:
- Terminal UI Standardization - Centralized terminal output utility for consistent CLI experience
- Skill Update Command - New
ai-devkit skill updatecommand for updating skills from registries- Update All Skills:
ai-devkit skill update- Updates all cached skill registries via git pull - Update Specific Registry:
ai-devkit skill update <registry-id>- Updates only the specified registry (e.g.,ai-devkit skill update anthropic/skills)
- Update All Skills:
- Module Resolution - Updated TypeScript configuration from Node16 to CommonJS for better compatibility
- Graceful Exit - Commands now properly exit with code 0 on successful completion
skill list- Added explicit process.exit(0) when no skills foundskill remove- Added explicit process.exit(0) after successful removal
- Custom Skill Registries - Support
skills.registriesin global~/.ai-devkit/.ai-devkit.jsonfor adding multiple registries that merge with defaults and override on conflicts. - Global Registry Reader - New global config reader for resolving custom registries in skill commands.
- Skill Registry Resolution - Skill commands now merge default and custom registries, with offline cache fallback when a registry URL is not configured.
- Memory Skill Template - New skill for integrating memory service capabilities into agent workflows
- Comprehensive Documentation - Added extensive documentation pages for:
- Getting Started guide
- Supported AI agents reference
- Development with AI DevKit
- Debug workflows
- Understanding existing code
- Memory service usage
- Skills management
- Updated base template for all environments
- Skill Management - Centralized registry for managing Agent Skills across projects
- One-Command Installation:
ai-devkit skill add <registry>/<repo> <skill-name> - Local Cache: Skills stored in
~/.ai-devkit/skills/to avoid duplication - Symlink-First Strategy: Symlinks with automatic copy fallback for Windows
- Multi-Environment Support: Works with Cursor, Claude Code, Codex, OpenCode, and Antigravity
- CLI Commands:
ai-devkit skill add <registry>/<repo> <skill-name>- Install a skill from registryai-devkit skill list- List all installed skills with sourcesai-devkit skill remove <skill-name>- Remove skill from project
- Features:
- Centralized registry file (
skills/registry.json) with verified repositories - Automatic
.ai-devkit.jsoncreation if missing - Environment filtering (only shows/uses environments with skill support)
- Git repository caching for efficient reuse across projects
- Validation for registry IDs and skill names (follows Agent Skills spec)
- Centralized registry file (
- One-Command Installation:
- Knowledge Memory Service (
packages/memory) - A lightweight MCP-based memory service for AI agents- Store and retrieve actionable knowledge using SQLite with FTS5 full-text search
- Core Features:
- 🔍 Full-Text Search - FTS5 with BM25 ranking
- 🏷️ Tag-Based Filtering - Boost results by contextTags
- 📁 Scoped Knowledge - global, project, or repo-specific rules
- 🔄 Deduplication - Prevents duplicate content
- CLI Integration: New
memorycommand familyai-devkit memory store- Store new knowledge itemsai-devkit memory search- Search for relevant knowledge
- Global Setup Command - New
ai-devkit setup --globalcommand for installing commands globally- Copy AI DevKit commands to global environment folders
- Support for Antigravity (
~/.gemini/antigravity/global_workflows/) and Codex (~/.codex/prompts/) - Interactive environment selection with only global-capable environments shown
- Overwrite prompts for existing global commands
- Cross-platform support using
os.homedir()andpath.join()
- Antigravity Support - Added support for Google Antigravity
- New Slash Command -
/simplify-implementationfor analyzing and simplifying existing implementations
- Dynamic TOML Generation - Refactored TemplateManager to dynamically generate
.tomlfiles from.mdfiles at runtime
- Fixed Gemini CLI integration #3
- Added test for TemplateManager.ts
- Fixed Github Copilot integration #4
- Multi-Environment Setup - Support for 10 AI development environments
- Interactive environment selection with multi-choice prompts
- Support for Cursor, Claude Code, GitHub Copilot, Google Gemini, OpenAI Codex, Windsurf, KiloCode, AMP, OpenCode, and Roo Code
- Unified template structure with AGENTS.md files for all environments
- Environment-specific command directories and configuration files
- Override protection with confirmation prompts for existing environments
- Config persistence storing selected environments array
- Breaking Changes - Removed legacy single-environment support for cleaner API
- Renamed
EnvironmentIdtoEnvironmentCodethroughout codebase - Removed legacy
Environmenttype union (cursor | claude | both) - Updated config schema to use
environments: EnvironmentCode[] - All environments now use standardized AGENTS.md context files
- Renamed
- Testing Infrastructure - Complete test suite implementation
- Architecture - Modular design improvements
/debug- Structured assistant for clarifying issues, analyzing options, and agreeing on a fix plan before coding/capture-knowledge- Analyze and explain how code works from any entry point- Supports file, folder, function, and API endpoint analysis
- Recursive dependency analysis with configurable depth (max: 3)
- Automatic generation of mermaid diagrams (flowcharts, sequence, architecture, class diagrams)
- Knowledge capture documentation saved to
docs/ai/implementation/knowledge-{feature-name}.md - Visual dependency tree and component relationship mapping
- Includes error handling, performance considerations, and improvement suggestions
- Eight slash commands for Cursor and Claude Code:
/new-requirement- Complete guided workflow from requirements to PR/MR creation/code-review- Structured local code reviews/execute-plan- Walk feature plans task-by-task/writing-test- Generate tests with guidance for 100% coverage/update-planning- Reconcile progress with planning docs/check-implementation- Compare implementation with design/review-design- Review system design and architecture/review-requirements- Review and summarize requirements
- Claude workspace configuration file (
CLAUDE.md) - Cursor rules file (
ai-devkit.md) - Design documentation requirements for mermaid diagrams (architecture and data flow)
- Initial release of AI DevKit CLI
- Interactive
initcommand for project initialization - Support for Cursor and Claude Code environments
- Seven phase templates: requirements, design, planning, implementation, testing, deployment, monitoring
phasecommand for adding individual phases- Configuration management with
.ai-devkit.json - Template overwrite prompts for existing files
- Comprehensive documentation and README
- TypeScript support with full type definitions
- Cursor rules in
.cursor/rules/directory - Cursor slash commands as individual Markdown files in
.cursor/commands/ - Claude Code workspace configuration in
CLAUDE.md
- Interactive prompts with Inquirer
- Flag-based overrides for automation
- Markdown templates with YAML frontmatter
- Cursor rules and slash commands generation
- Claude Code workspace configuration
- State tracking for initialized phases