v0.2.7

Release Notes - v0.2.7

Summary

Bug fix release focusing on workflow reliability improvements for phase advancement and ledger management.

Bug Fixes

• Fixed manual testing tasks in tasks.md - Updated technical-architect agent template to prevent manual testing tasks from being generated in tasks.md. Manual tests now only appear in testing.md, resolving phase advancement failures caused by unchecked manual test tasks.

• Fixed initLedger overwrite - Resolved critical data loss bug where initLedger() overwrote existing ledger during re-initialization. Ledgers are now preserved unless corrupted.

• Fixed repair command ledger reconstruction - Extended repair command to detect and reconstruct missing ledger entries from feature folders in todo/ and done/ directories.

Improvements

• Performance and type safety optimizations for JS/TS operations

v0.2.6

Release Notes - v0.2.6

Release Date: 2025-12-24

Highlights

This release introduces a completely redesigned skill injection system that ensures agents always receive the context they need, plus powerful new agent templates and skill placeholder features.

What's New

Orchestrator-Driven Skill Injection

The skill loading system has been completely redesigned for reliability. Previously, agents loaded skills via "First Action" instructions which was unreliable. Now, slash commands pre-load phase skills and inject them directly into Task prompts, guaranteeing agents always have the required context when they start working.

Agent Templates

Complete agent templates feature - agents can now be defined with reusable templates, making it easier to create consistent agent configurations across your project.

Skill Placeholder Injection

New skill placeholder injection system allows dynamic skill content to be injected into agent prompts, enabling more flexible and context-aware agent behavior.

Improvements

• General code cleanup and issue resolution

v0.2.5

• Portable Skills & Agents: Removed hardcoded output paths, making resources work across different project locations

v0.2.4

• Renamed Skills: refinement-questions → refinement-product-requirements, refinement-spec-writer → refinement-technical-specs for clarity
• Enhanced Technical Skill: Now includes codebase exploration and technical Q&A phases

v0.2.3

• Simplified Workflow: Reverted type-specific workflows back to single unified flow for all item types
• Phase Detection Fix: Corrected edge cases for review and testing failure handling

v0.2.2

feat(sync): implement auto-discovery and improved change tracking

Replace hardcoded resource manifest with runtime directory scanning
to automatically detect agents, skills, and commands. Framework files
in .nextai/ now always update without --force flag.

Changes:
- Add scanResourcesFromPackage() for auto-discovery
- Implement ResourceStats interface for change tracking
- Show detailed sync output: "Commands: 13 (1 new, 2 updated)"
- Remove force checks from .nextai/ copy functions
- Add deprecated resource removal for .nextai/ only
- Add "Upgrading NextAI" section to README
- Update project documentation

v0.2.1

🐛 Bug fixes

• fix(sync): correct agents directory path

v0.2.0

Features

• Metrics System - Track feature lifecycle performance with phase durations, review iterations, and aggregated statistics. Data stored in .nextai/metrics/

• Spec Change Detection - Testing-investigator now detects when test failures indicate spec changes vs bugs. Interactive approval flow resets features to refinement phase when needed

v0.1.1

Tests learned cross-platform manners, line endings no longer cause drama

v0.1.0

Rename package to frontztech/nextai-dev for npm publishing