# Changelog

All notable changes to this project will be documented in this file.

The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/),
and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).

## [0.2.0] - 2026-08-12

Naming unification release: the project is now consistently **`harness-ai-kit`**, plus 11 additional curated skills.

### Changed

- **BREAKING — the canonical name is now `harness-ai-kit` throughout.** The Python package (`harness_ai_kit`), primary command (`harness-ai-kit`), project manifest (`harness-ai-kit.yml`), lockfile (`harness-ai-kit.lock`), config directory (`~/.harness-ai-kit/`), runtime skill bundle root (`harness-ai-kit-skills`), and product env var (`HARNESS_AI_KIT_PRODUCT`) were all unified from the earlier `ai-kit` naming.
  - The short command **`ai-kit` is retained as an alias** and keeps working (`ai-kit sync` == `harness-ai-kit sync`).
  - **Migration for early adopters:** rename `ai-kit.yml` → `harness-ai-kit.yml` and `ai-kit.lock` → `harness-ai-kit.lock` in your project; move `~/.ai-kit/` → `~/.harness-ai-kit/`; then re-run `harness-ai-kit sync`. Skill inject markers (`<!-- ai-kit:inject ... -->`) become `<!-- harness-ai-kit:inject ... -->` and are refreshed automatically on the next sync.

### Added

- **11 additional curated skills** (34 total): AI engineering methodology (spec-driven dev, agent engineering, eval-driven agent, ai-kit miner, tech-debt ops), patent & document authoring (6), and testing/quality (web deep acceptance, web E2E, tech-fit eval)
- **Skill catalog** (`CATALOG.md`) and **usage scenarios** (`docs/usage-scenarios.md`) covering the skill+SDD and loop+runtime models

## [0.1.0] - 2026-08-10

Initial public release of harness-ai-kit — the package manager for AI agent assets.

### Added

- **CLI core**: `harness-ai-kit` command with init, add, sync, lock, resolve, graph, list, doctor, verify, remove, search subcommands
- **Unified asset model**: six asset kinds (skill / cli / mcp / plugin / hook / subagent) sharing a single typed dependency graph
- **Dependency solver**: SAT-based resolution via resolvelib with feature flags and scope overrides
- **Lockfile**: `harness-ai-kit.lock` with SHA-256 content integrity verification
- **Multi-runtime sync**: Codex, Claude Code, Cursor, Kiro — project-level or global install
- **GitHub direct install**: install skills from any GitHub repo without a private registry
- **Atomic install**: staging-based install with automatic rollback on failure
- **Offline mode**: cache-backed offline install for air-gapped environments
- **23 curated skills**:
  - 9 expert bases: MySQL, PostgreSQL, Redis, MongoDB, Kafka, RabbitMQ, Oracle, NL2SQL, Git workflow
  - 7 diagnostic playbooks: MySQL deadlock / slow query / replication, container OOM, K8s CrashLoop / node pressure, network port unreachability
  - 6 utilities: registry mirror strategy, goal execution, markitdown, work convert, work export, post-task skill miner
  - 1 flagship: infra system environment ops
- **CI pipeline**: lint, import check, CLI smoke test, skill validation, sensitive-scan, secret-scan (Python 3.10 / 3.11 / 3.12 matrix)
- **Community scaffolding**: CONTRIBUTING, CODE_OF_CONDUCT, issue templates, PR template
- **Bilingual docs**: English README + Chinese README, quickstart, concepts, skill authoring guide, asset map

### Security

- No internal network references, credentials, or proprietary endpoints in shipped code
- CI sensitive-scan blocks internal domain/IP patterns on every push