Plugin suite + bundled MCP servers for Claude Code. Full delivery lifecycle: Agile pipeline with multi-model AI review, project bootstrap, documentation generation, codebase audits, performance optimization, community workflows. Includes hex-line (hash-verified editing), hex-graph (code knowledge graph), and hex-ssh (remote SSH) MCP servers.
AI Agentsagile-workflowsai-agentsanthropicclaude-aiclaude-code
Installation
# Add to your Claude Code skills
git clone https://github.com/levnikolaevich/claude-code-skills
README.md
Claude Code Skills
7 plugins. One marketplace. Install only what you need and automate your full delivery workflow —
from project bootstrap to code audit to production quality gates.
Works standalone or as a complete Agile pipeline.
[!TIP]
Multi-Model AI Review — Delegate code & story reviews to Codex and Gemini agents running in parallel, with automatic fallback to Claude Opus. Ship faster with 3x review coverage.
Browse and discover individual skills at skills.sh.
[!NOTE]
skills.sh is a showcase only. Skills depend on shared resources (shared/ directory) that are not copied by npx skills add. Use /plugin marketplace add and /plugin install for a working installation.
ln-010-dev-environment-setup # Set up agents, MCP, sync configs
ln-620-codebase-auditor # Audit your code for issues
ln-100-documents-pipeline # Generate documentation
Full Agile workflow (Linear or File Mode — auto-detected):
Manual step-by-step (if you prefer control over each stage):
ln-400-story-executor # Execute Story tasks
ln-500-story-quality-gate # Quality gate + test planning
Workflow
ln-010-dev-environment-setup # 0. Set up dev environment (once)
↓
ln-100-documents-pipeline # 1. Documentation
↓
ln-200-scope-decomposer # 2. Scope -> Epics -> Stories
↓
ln-1000-pipeline-orchestrator # 3. Full artifact-driven pipeline: 300 → 310 → 400 → 500 → Done
MCP Servers (Optional)
Bundled MCP servers extend agent capabilities — hash-verified editing, code intelligence, and remote access. All skills work without MCP (fallback to built-in tools), but MCP servers improve accuracy and save tokens.
Bundled servers
| Server | What it does | Tools | Docs |
|--------|-------------|-------|------|
| hex-line-mcp | Every line carries a content hash — edits prove the agent sees current content. Prevents stale-context corruption. Includes validation hooks. | 9 | README · npm |
| hex-graph-mcp | Indexes codebases into a deterministic SQLite graph with framework-aware overlays, capability-first quality tooling, optional SCIP interop, and architecture/reference analysis. | 14 | README · npm |
| hex-ssh-mcp | Hash-verified remote file editing and SFTP transfer over SSH. Normalized output for minimal token usage. | 8 | README · npm |
Deterministic scope rule: hex-line and hex-graph keep path as the project anchor. In normal use the agent fills it automatically from the active file or project root, so users usually do not need to type it manually. hex-ssh runs on Windows/macOS/Linux hosts; remote shell tools stay POSIX-oriented, while SFTP transfers support platform-aware remote paths.
| Server | Purpose | API Key | Used by |
|--------|---------|---------|---------|
| Context7 | Library docs, APIs, migration guides | Optional (dashboard) | ln-310, ln-511, ln-640+ |
| Ref | Standards, RFCs, best practices | Required (ref.tools/keys) | ln-310, ln-511, ln-640+ |
| Linear | Issue tracking (Agile workflow) | OAuth via browser | ln-300+, ln-400+, ln-500+ |
CLI setup:
# hex-line — hash-verified file editing (bundled)
npm i -g @levnikolaevich/hex-line-mcp
claude mcp add -s user hex-line -- hex-line-mcp
# hex-ssh — token-efficient SSH with hash verification (bundled)
npm i -g @levnikolaevich/hex-ssh-mcp
claude mcp add -s user hex-ssh -- hex-ssh-mcp
# hex-graph — code knowledge graph (bundled)
npm i -g @levnikolaevich/hex-graph-mcp
claude mcp add -s user hex-graph -- hex-graph-mcp
# Context7 — library documentation (HTTP, optional API key)
claude mcp add -s user --transport http --header "CONTEXT7_API_KEY: YOUR_KEY" context7 https://mcp.context7.com/mcp
# Ref — standards & best practices (HTTP, API key required)
claude mcp add -s user --transport http --header "x-ref-api-key: YOUR_KEY" Ref https://api.ref.tools/mcp
# Linear — issue tracking (HTTP, OAuth via browser)
claude mcp add -s user --transport http linear-server https://mcp.linear.app/mcp
Agent steering
MCP servers can be installed correctly and still lose to built-ins in practice. hex-line-mcp keeps Claude aligned through one output style and three Claude hook events:
| Mechanism | How it works |
|-----------|-------------|
| Output style | Injected into system prompt — maps built-in tools to MCP equivalents (Read → hex-line read_file, Edit → hex-line edit_file) |
| SessionStart hook | Injects a compact bootstrap hint and defers to the active hex-line output style when present |
| PreToolUse hook | Hard-redirects project text Read/Edit/Write/Grep/Glob, redirects project file-inspection Bash commands, blocks dangerous commands |
| PostToolUse hook | Filters only verbose Bash output (50+ lines), keeping first 15 + last 15 lines after normalization and dedupe |
Hooks and output style auto-sync on hex-line-mcp startup. First run after install performs the initial sync automatically.
AI Review Models (Optional)
Multi-model review uses external AI agents (Codex + Gemini) for parallel code/story analysis. Both agents run simultaneously with automatic fallback to Claude Opus if unavailable.
| Model | CLI | Version | Used by | Settings |
|-------|-----|---------|---------|----------|
| Codex | codex | gpt-5.4 | ln-310, ln-510, ln-813 | --json --full-auto (read-only, internet access) |
| Gemini | gemini | Auto (Gemini 3) | ln-310, ln-510, ln-813 | --yolo (sandbox, auto-approve, auto model selection) |
Review Workflow:
Parallel Execution — Both agents run simultaneously (background tasks)
Critical Verification — Claude validates each suggestion (AGREE/DISAGREE/UNCERTAIN)
Debate Protocol — Challenge rounds (max 2) for controversial findings
Filtering — Only high-confidence (≥90%), high-impact (>2%) suggestions surface
Fallback — Self-Review (Claude Opus) if agents unavailable
The agent harness performance optimization system. Skills, instincts, memory, security, and research-first development for Claude Code, Codex, Opencode, Cursor and beyond.
