Release: v0.6.7

New Features

• Cursor Editor support: Install skills to Cursor with --cursor option

  npx agentic-code skills --cursor           # User scope (~/.cursor/skills/)
  npx agentic-code skills --cursor --project # Project scope (./.cursor/skills/)

  Note: Skills feature in Cursor requires Nightly release channel.

Changes

• Rename install-codex-skills.js to install-skills.js for consistency
• Update README to prioritize Cursor over Codex (reflecting popularity)
• Existing --codex option continues to work unchanged

Release: v0.6.6

Improvements

• File count estimation process: Added mandatory investigation steps for scale determination (identify entry points, trace dependencies, include test files, list affected paths explicitly)
• Scale confidence: Introduced confirmed/provisional indicator to show certainty of scale determination
• Scope dependencies: Added documentation of factors that could change scale based on user answers

Release: v0.6.5

What's Changed

Features

• Add complexity gate for over-engineering detection in design review process
  • New complexity_level and complexity_rationale fields in Design Doc template
  • Complexity assessment check added to Design Doc completion conditions
  • Document reviewer now flags unjustified medium/high complexity as needs_revision

Full Changelog: v0.6.4...v0.6.5

Release: v0.6.4

What's Changed

Bug Fix

• Fix skills install command: npx agentic-code-install-skills was not working because npx looks for a package name, not a bin command within a package.

New Usage

npx agentic-code skills --codex
npx agentic-code skills --codex --project
npx agentic-code skills --path ./custom

Full Changelog: v0.6.3...v0.6.4

Release: v0.6.3

Fixed

• Convert skill descriptions from YAML folded style (>) to single-line format for better LLM compatibility

Release: v0.6.2

New Features

• Project-level skill installation: --codex --project installs skills to ./.codex/skills/ for per-project configuration
• Custom path support: --path <path> allows installing skills to any directory
• CODEX_HOME environment variable: Respects custom Codex home directory for user-scope installations

Improvements

• Improved skill descriptions with action verbs and trigger words for better discoverability
• Changed YAML multiline style from literal (|) to folded (>) for single-line display compatibility

Usage

# User scope (default)
npx agentic-code-install-skills --codex

# Project scope
npx agentic-code-install-skills --codex --project

# Custom path
npx agentic-code-install-skills --path ./custom/skills

Release: v0.6.1

What's Changed

Fix: Add early exit check for operational tasks

Tasks that don't require reading code files to decide HOW to execute (e.g., git operations, simple commands) now bypass task-analysis.md and execute directly.

This prevents unnecessary workflow overhead for simple operational tasks like:

• Git operations (checkout, pull, push, branch, etc.)
• Environment checks
• Simple file operations

Full Changelog: v0.6.0...v0.6.1

Release: v0.6.0

Skills Migration (Codex Compatible)

Migrated .agents/rules/ to .agents/skills/ using the Codex Skills format.

Changes

• Skills format: Each skill now has a SKILL.md with YAML frontmatter (name, description)
• Language references: Language-specific content moved to references/ subdirectory (e.g., references/typescript.md)
• Config renamed: task-rule-matrix.yaml → task-skills-matrix.yaml
• CLI simplified: Removed --lang option (no longer needed)

New Command

Install skills globally for Codex:

npx agentic-code-install-skills --codex
# Installs to ~/.codex/skills/agentic-code/

Terminology Fix

Renamed "Plan Injection" to "Task Tracking" to avoid confusion between LLM's internal task management and Work Plan documents (docs/plans/).

Breaking Changes

• .agents/rules/ directory renamed to .agents/skills/
• Skill file paths changed (e.g., rules/core/metacognition.md → skills/metacognition/SKILL.md)

Release v0.5.2

What's Changed

Testing Rules Enhancement

Added positive criteria to help AI detect test antipatterns without triggering the pink elephant problem:

• Tautology test detection: Added "Expected value ≠ mock return value" criterion to Literal expectations in general/testing.md
• Private method test detection: Added method call example (service as any).validate(input) to typescript/testing.md

These criteria enable AI to verify test validity through positive rules rather than antipattern lists.

Release v0.5.1

What's Changed

• Add files field to package.json for npm packaging
• Update installation commands to use npx agentic-code

npm Installation

This release enables npm registry publishing. After this release:

npx agentic-code my-project

Full Changelog: v0.5.0...v0.5.1