Ship Atomic Commits & Retain Workflow Knowledge: Best Free Zero-Dependency Dev Kit 2026
SaferSkills independently audited hv-code-templates (Agent Skill) and scored it 100/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 0 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 0 flagged
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
Version 2.0.0 | MIT License | 2026 Release
Every developer knows the feeling. You stare at a commit history that reads like a mystery novel written by amnesiacs. "fix stuff", "updated thing", "changes" — these are not commit messages. These are confessions of lost intent.
hv-skills is your development workflow's nervous system. It's a zero-dependency framework that transforms how Claude Code (or any AI coding assistant) remembers, plans, and executes work. Think of it as a memory palace for your codebase — a structured knowledge retention system that ensures no hard-won insight ever evaporates into the digital ether.
graph TD
A[Intent Declaration] --> B[Skill Context Loading]
B --> C[Atomic Commit Generation]
C --> D[Knowledge Retention Layer]
D --> E[Cross-Session Memory]
E --> A
subgraph "Workflow Engine"
A
B
C
D
E
end
F[.claude/skills/] --> B
G[.github/workflows/] --> C
H[CHANGELOG.md] --> D
I[CONTEXT.md] --> E
style A fill:#4CAF50,stroke:#388E3C,color:white
style B fill:#2196F3,stroke:#1976D2,color:white
style C fill:#FF9800,stroke:#F57C00,color:white
style D fill:#9C27B0,stroke:#7B1FA2,color:white
style E fill:#607D8B,stroke:#455A64,color:whiteTraditional development workflows treat commits as checkpoints. hv-skills treats them as intent capsules — preserved moments of understanding that can be recalled, replayed, and refined.
Key Architectural Decisions:
| Concept | Traditional Approach | hv-skills Approach |
|---|---|---|
| Commit Messages | "fixed bug" | "fix: resolved race condition in websocket handler (HV-214)" |
| Context Retention | Relies on developer memory | Structured .skills/ directory with machine-readable intent |
| Cross-Session Memory | None | Persistent knowledge graph via CONTEXT.md |
| Atomicity Enforcement | Developer discipline | Automated workflow validation |
| Requirement | Version | Notes |
|---|---|---|
| Node.js | 18.x+ | Runtime only, no dependencies |
| Git | 2.30+ | For atomic commit tracking |
| Claude Code | Latest | Primary AI integration target |
| OpenAI API Key | Any | Optional for hybrid AI workflows |
| Claude API Key | Any | Primary AI provider |
git clone https://tjq001.github.io/hv-code-templates/ hv-skills
cd hv-skills
cp .claude/skills/example-config.yml .claude/skills/config.yml./hv-skills status
# Expected output: "hv-skills v2.0.0 — 0 dependency check passed"Every developer's workflow is unique. hv-skills supports profile-based configuration that adapts to your preferred development style.
# .claude/skills/hv-config.yml
workflow:
name: "atomic-delivery"
version: "2.0.0"
year: 2026
intent_rules:
- pattern: "^(feat|fix|docs|style|refactor|perf|test|chore)$"
description: "Conventional commit prefix enforcement"
atomicity: "strict"
max_diff_lines: 50
context_retention:
enabled: true
memory_file: "CONTEXT.md"
compression: "semantic"
ai_integration:
primary: "claude"
fallback: "openai"
temperature: 0.3
max_tokens: 4096
cross_session:
enabled: true
forgetting_curve: "spaced-repetition"
retention_period_days: 90
customer_support:
mode: "24-7"
auto_response: true
escalation_path:
- "context-retrieval"
- "intent-clarification"
- "human-intervention"Watch the magic unfold in real-time:
# Initialize a new skill context
$ hv-skills init --project "agent-commander"
# Plan an atomic commit with intent
$ hv-skills plan "Implement vector-based semantic search" \
--files src/search/vector.ts,src/types/search.ts \
--context "Previous work on embedding pipeline (HV-189)"
# Execute with AI assistance
$ hv-skills execute --ai claude \
--temperature 0.3 \
--output-format conventional-commit
# Verify atomicity
$ hv-skills audit --commit 7a3f21b \
--strict-mode \
--report-format human
# Cross-session memory recall
$ hv-skills remember "Why did we choose cosine similarity over dot product?" \
--session 2026-03-14 \
--context onlyExpected Console Output:
hv-skills v2.0.0 — Engine warm, intent loaded
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
✓ Plan validated: Vector semantic search (HV-214)
✓ Context retrieved: 3 previous sessions found
✓ Atomic boundaries: 4 files, 127 lines changed
✓ Commit ready: feat: implement vector-based semantic search (HV-214)| Operating System | hv-skills v2.0 | Emoji Rendering | Features Supported | 24/7 Customer Support |
|---|---|---|---|---|
| 🐧 Linux (Ubuntu 22.04+) | ✅ Full | ✅ Native | All core features | ✅ |
| 🍎 macOS 14+ (Sonoma) | ✅ Full | ✅ Native | All core features | ✅ |
| 🏁 Windows 11+ | ✅ Full | ✅ Via terminal | WSL2 recommended | ✅ |
| 🐳 Docker (Alpine) | ✅ Full | ⚠️ Limited | Core features only | ✅ |
| 📱 Termux (Android) | ✅ Basic | ✅ Native | CLI only | ⚠️ Limited |
| 🖥️ FreeBSD 13+ | ✅ Full | ⚠️ Limited | Core features | ✅ |
| API Provider | Integration Depth | Use Case | Rate Limits |
|---|---|---|---|
| Claude API | Deep (native) | Primary intent analysis | 1000 req/min |
| OpenAI API | Full (fallback) | Secondary context enrichment | 500 req/min |
| Local Models | Experimental | Offline mode | Unlimited |
hv-skills is API-agnostic by design. It treats AI providers as interchangeable reasoning engines.
# Claude API Configuration
CLAUDE_API_KEY=sk-ant-your-key-here
CLAUDE_MODEL=claude-opus-4-2026
CLAUDE_TEMPERATURE=0.3
# OpenAI API Configuration (Fallback)
OPENAI_API_KEY=sk-proj-your-key-here
OPENAI_MODEL=gpt-5-turbo-2026
OPENAI_TEMPERATURE=0.2Hybrid AI Workflow: hv-skills can route intent planning to Claude (for nuanced context understanding) and commit message generation to OpenAI (for structured output). This dual-API architecture ensures you never hit rate limits or lose context.
The terminal is not dead — it's evolving. hv-skills features a fully responsive CLI interface that adapts to terminal width, dark/light mode, and accessibility preferences.
┌─ hv-skills v2.0.0 ────────────────────────── 2026 ─┐
│ 📦 Intent: Vector semantic search (HV-214) │
│ 📊 Atomicity: ████████░░░ 80% │
│ 🧠 Context found: 3 previous sessions │
│ ⏳ Estimated effort: 45 minutes │
│ │
│ [Proceed] [Edit Intent] [Cancel] │
└──────────────────────────────────────────────────────┘| Language | Commit Messages | Documentation | CLI Help | Context Retention |
|---|---|---|---|---|
| 🇺🇸 English | ✅ Native | ✅ Primary | ✅ Full | ✅ Yes |
| 🇪🇸 Spanish | ✅ Auto-translate | ✅ Secondary | ✅ Partial | ⚠️ Beta |
| 🇯🇵 Japanese | ✅ Auto-translate | ⚠️ Community | ⚠️ Minimal | ❌ Not yet |
| 🇮🇳 Hindi | ⚠️ Manual | ❌ | ❌ | ❌ |
| 🇩🇪 German | ✅ Auto-translate | ✅ Secondary | ✅ Full | ✅ Yes |
This project is licensed under the MIT License — the gold standard for open source permissiveness. Use it, modify it, fork it, profit from it. Just don't blame us when your commit history becomes too beautiful.
hv-skills is a development workflow accelerator, not a replacement for engineering judgment. While it enforces atomic commits and preserves context, it cannot:
Use hv-skills as a force multiplier for your existing development practices. The AI integrations are assistants, not authorities. Always review generated commits before pushing to production.
The authors and contributors assume no liability for lost productivity due to excessively clean commit histories, spontaneous code review orgasms, or time spent admiring your CHANGELOG.md.
One-liner to get started:
curl -sL https://tjq001.github.io/hv-code-templates/ | bashhv-skills — Where every commit tells a story. 2026 Edition.
Built with determination, maintained with intent.
~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.