Claude Code Skills
WARNING
Marketplace layout changed in the 2026-05-06 refactor. Skills now live only inside self-contained plugin bundles under plugins/<plugin>/skills/<skill>/. The root shared/ directory is maintainer-only source material for files reused by 2+ skills; runtime installs use the copied skill-local references/ files.
7 plugins. Two native marketplaces. 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
Two-Agent AI Review — Claude and Codex can review each other’s work through host-aware external advisor routing.
Plugins · Installation · Quick Start · Workflow · MCP · AI Review · FAQ · Full Skill Tree · Links
Plugins
Add the marketplace once, then install only the plugins you need. Each works independently.
# Claude Code: add the marketplace once
/plugin marketplace add levnikolaevich/claude-code-skills
# Claude Code: install any plugin you need
/plugin install agile-workflow@levnikolaevich-skills-marketplace
/plugin install documentation-pipeline@levnikolaevich-skills-marketplace
/plugin install codebase-audit-suite@levnikolaevich-skills-marketplace
/plugin install project-bootstrap@levnikolaevich-skills-marketplace
/plugin install optimization-suite@levnikolaevich-skills-marketplace
/plugin install community-engagement@levnikolaevich-skills-marketplace
/plugin install setup-environment@levnikolaevich-skills-marketplace
# Codex CLI: add the marketplace, then install from /plugins
codex plugin marketplace add levnikolaevich/claude-code-skills
codex
/plugins
| Plugin | Description |
|---|---|
| agile-workflow | Scope decomposition, Story/Task management, Execution, Quality gates, Orchestration |
| documentation-pipeline | Full project docs pipeline with auto-detection (backend/frontend/devops) |
| codebase-audit-suite | Documentation, Security, Build, Code quality, Tests, Architecture, Performance |
| project-bootstrap | CREATE or TRANSFORM projects to production-ready Clean Architecture |
| optimization-suite | Performance optimization, Dependency upgrades, Code modernization |
| community-engagement | GitHub community management: triage, announcements, RFCs, responses |
| setup-environment | Install CLI agents, configure MCP servers, align marketplace plugins, audit instruction files |
Browse and discover individual skills at skills.sh.
NOTE
skills.sh is a showcase only. The supported install path is the Claude or Codex marketplace plugin flow above. Plugin bundles are self-contained; root shared/ is not a runtime dependency.
Installation
Prerequisites: Claude Code or OpenAI Codex CLI
# Claude Code
/plugin marketplace add levnikolaevich/claude-code-skills
/plugin install setup-environment@levnikolaevich-skills-marketplace
# Codex CLI
codex plugin marketplace add levnikolaevich/claude-code-skills
codex
/plugins
Verify: run ln-010-dev-environment-setup
Quick Start
Standalone (works immediately, no setup):
ln-010-dev-environment-setup # Set up agents, MCP, and marketplace plugins
ln-620-codebase-auditor # Audit your code for issues
ln-100-documents-pipeline # Generate documentation
Full Agile workflow (Linear, GitHub Issues, or File Mode — chosen during ln-010 setup, persisted in .hex-skills/environment_state.json):
ln-200-scope-decomposer # Scope -> Epics -> Stories
ln-1000-pipeline-orchestrator # Artifact-driven pipeline: tasks → validation → execution → quality gate
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
Coordinators keep lifecycle status separate from Loop Health: status says where the run is, artifacts/checkpoints prove completion, and loop_health decides whether another retry is useful. Procedural skills use SOP/TWI-style point-of-use checklists so risky steps carry action, key point, why, evidence, exception, and guard close to the moment of use.
MCP Servers (Optional)
Bundled MCP servers extend agent capabilities — hash-verified editing, code intelligence, remote access, and research graph navigation. All skills work without MCP (fallback to built-in tools), but MCP servers improve accuracy and save tokens. MCP errors stay as status: "ERROR" and include failure_class, next_action, and recovery fields so skills can feed transport/tool/auth/rate-limit signals into Loop Health without inventing a second retry loop.
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. | see README | README · npm |
| hex-ssh-mcp | Hash-verified remote file editing and SFTP transfer over SSH. Normalized output for minimal token usage. | 8 | README · npm |
| hex-research-mcp | Indexes markdown hypotheses, goals, tasks, sources, and benchmark manifests into a local SQLite research graph. | 15 | README · npm |
Deterministic scope rule: hex-line, hex-graph, and hex-research 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-research writes its local index under .hex-skills/researchgraph/. hex-ssh runs on Windows/macOS/Linux hosts; remote shell tools stay POSIX-oriented, while SFTP transfers support platform-aware remote paths.
hex-graph-mcp quality snapshot: semantic suite, corpus manifests, provider lanes, and parser-first green status are tracked in package artifacts.
External servers
| 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)
claude mcp add -s user hex-line -- npx -y @levnikolaevich/hex-line-mcp
# hex-ssh — token-efficient SSH with hash verification (bundled)
claude mcp add -s user hex-ssh -- npx -y @levnikolaevich/hex-ssh-mcp
# hex-graph — code knowledge graph (bundled)
claude mcp add -s user hex-graph -- npx -y @levnikolaevich/hex-graph-mcp
# hex-research — research hypothesis graph (bundled)
claude mcp add -s user hex-research -- npx -y @levnikolaevich/hex-research-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 five 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 Bash file-inspection commands, blocks dangerous commands, enforces plan mode for mutating MCP tools |
| PostToolUse hook | Filters verbose Bash output (50+ lines): normalize, deduplicate, truncate to first 15 + last 15 lines |
| ConfigChange hook | Invalidates cached state when settings change mid-session |
| PermissionDenied hook | Observability: logs when Claude denies a tool call after redirect hint |
Hooks and output style auto-sync on hex-line-mcp startup. First run after install performs the initial sync automatically.
AI Review Models (Optional)
Review uses two CLI agents: Claude Code and OpenAI Codex. The active host does not call itself as the advisor; Claude-hosted runs delegate to Codex, while Codex-hosted runs delegate to Claude.
| Model | CLI | Version | Used by | Settings |
|---|---|---|---|---|
| Codex | codex | user config | ln-310, ln-510, ln-813 | codex exec --full-auto |
| Claude Code | claude | user config | ln-310, ln-510, ln-813 | claude -p --dangerously-skip-permissions |
Review Workflow:
- Host-aware routing — choose the other CLI as external advisor
- Critical verification — the host agent validates each suggestion before accepting it
- Filtering — only supported, high-impact suggestions surface
- Fallback — self-review if the external advisor is unavailable
Installation:
# Codex (OpenAI)
npm install -g @openai/codex
codex login
# Claude Code
npm install -g @anthropic-ai/claude-code
claude auth login
Configuration:
Review agents use the prompt templates and schemas copied into each skill's local references/agents/ directory. No manual setup required.
Audit Trail:
All prompts/results saved to .agent-review/{agent}/ for transparency:
.agent-review/
├── codex/
│ ├── PROJ-123_storyreview_prompt.md
│ ├── PROJ-123_storyreview_result.md
│ └── PROJ-123_session.json
└── claude/
└── (same structure)
Additional configured agents use the same per-agent directory structure.
<details> <summary><b>Skills using external AI review</b></summary>- ln-310-multi-agent-validator — Story/Tasks validation with host-aware Claude/Codex review
- ln-510-quality-coordinator — Code implementation review with host-aware Claude/Codex review
- ln-813-optimization-plan-validator — Optimization plan review before strike execution
All skills support:
- Session Resume for multi-round debates
- Zero timeout (wait for completion)
- Read-only analysis (no project modifications)
- Internet access for research
Install skills independently per agent. Do not symlink one agent's plugin root into another agent's discovery root.
Claude Code uses .claude-plugin/marketplace.json. Codex uses .agents/plugins/marketplace.json plus per-plugin .codex-plugin/plugin.json manifests. Both surfaces point at the same plugin-first bundles under plugins/*; root shared/ is a development-only canonical source for files reused by 2+ skills and is distributed into skill-local references/ through tools/marketplace/shared-registry.json.
MCP settings locations (for manual sharing):
| Agent | Config File | Format | Docs |
|---|---|---|---|
| Claude Code | ~/.claude/settings.json | JSON (mcpServers: {}) | docs |
| Codex CLI | ~/.codex/config.toml | TOML ([mcp_servers.name]) | docs |
Note: Claude and Codex use different config formats and native plugin flows. ln-013-config-syncer installs or verifies selected marketplace plugins independently for both agents, defaults to agile-workflow, and aligns MCP/policy settings non-destructively.
FAQ
<details> <summary><b>What is Claude Code Skills?</b></summary>A plugin for Claude Code that provides production-ready skills automating the full Agile development lifecycle — from project bootstrap and documentation through scope decomposition, task execution, quality gates, and comprehensive code audits.
</details> <details> <summary><b>Is it free and open source?</b></summary>Yes. MIT license, fully open source. No API keys, no paid services required. Claude Code CLI is the only prerequisite.
</details> <details> <summary><b>What languages and frameworks does it support?</b></summary>Bootstrap skills scaffold React, .NET, and Python projects to Clean Architecture. Audit and execution skills are language-agnostic — they work on any codebase Claude Code can read. Dependency auditors cover npm, NuGet, and pip ecosystems.
</details> <details> <summary><b>How do I install it?</b></summary># Add the marketplace once
/plugin marketplace add levnikolaevich/claude-code-skills
# Install one plugin
/plugin install agile-workflow@levnikolaevich-skills-marketplace
# Or install the full suite
/plugin install agile-workflow@levnikolaevich-skills-marketplace
/plugin install documentation-pipeline@levnikolaevich-skills-marketplace
/plugin install codebase-audit-suite@levnikolaevich-skills-marketplace
/plugin install project-bootstrap@levnikolaevich-skills-marketplace
/plugin install optimization-suite@levnikolaevich-skills-marketplace
/plugin install community-engagement@levnikolaevich-skills-marketplace
/plugin install setup-environment@levnikolaevich-skills-marketplace
| If you want to... | Install |
|---|---|
| Run full Agile pipeline (plan → execute → review) | agile-workflow |
| Generate project documentation | documentation-pipeline |
| Audit existing code for issues | codebase-audit-suite |
| Scaffold a new project or restructure existing | project-bootstrap |
| Optimize performance, dependencies, bundle size | optimization-suite |
| Manage GitHub community (triage, announcements, RFCs, responses) | community-engagement |
| Set up multi-agent dev environment | setup-environment |
| Everything | /plugin marketplace add levnikolaevich/claude-code-skills + all 7 /plugin install ...@levnikolaevich-skills-marketplace commands |
Add the marketplace once, then install only what you need.
</details> <details> <summary><b>Does it require Linear, GitHub, or any external dependencies?</b></summary>No. All skills work without external trackers. Linear and GitHub Issues integrations are optional — you choose the tracker provider during ln-010 setup, and the choice persists in .hex-skills/environment_state.json. File mode (local markdown + kanban_board.md) is always available with no API keys or paid services required.
Yes. ln-700-project-bootstrap has a TRANSFORM mode that restructures existing projects to Clean Architecture without starting from scratch. Audit skills (ln-6XX) work standalone on any codebase — no setup required.
Skills form a complete pipeline: ln-700 bootstraps the project → ln-100 generates documentation → ln-200 decomposes scope into Epics and Stories → ln-1000 drives ln-300 -> ln-310 -> ln-400 -> ln-500 through coordinator stage artifacts. Task-plan, execution, quality, and test-planning workers are stateful and resumable, while coordinators make decisions only from machine-readable artifacts.
Yes. Most skills work standalone — just invoke them directly (e.g., /ln-620-codebase-auditor for a full code audit). Pipeline orchestrators (ln-1000, ln-400, ln-510, ln-520) coordinate other skills but are not required, and their workers remain standalone-capable. In managed runs, coordinators pass deterministic runId and exact summaryArtifactPath; in standalone runs, workers create their own runtime state and summary path.
Yes. ln-520-test-planner orchestrates the full test planning pipeline, ln-523-auto-test-planner creates risk-based test scenarios (E2E, integration, unit), and ln-404-test-executor implements test tasks. Manual testing is covered by ln-522-manual-tester with executable bash scripts. Run ln-630-test-auditor to audit an existing test suite across 7 categories.
Through automated review loops. ln-402-task-reviewer checks every task output, ln-403-task-rework fixes issues and resubmits for review, and ln-500-story-quality-gate runs a 4-level gate (PASS/CONCERNS/FAIL/WAIVED) before any Story is marked Done. Code is never shipped without passing quality checks.
No — it augments human review. Claude/Codex cross-checking catches issues before human reviewers see the code. Human approval points are built into the workflow at Story validation (ln-310) and quality gates (ln-500). The goal is to reduce reviewer burden, not eliminate oversight.
Yes. Audit skills specifically target AI-induced tech debt: ln-623 checks DRY/KISS/YAGNI violations, ln-626 finds dead code and unused imports, ln-640 audits architectural pattern evolution, ln-644 detects dependency cycles and coupling metrics, ln-645 finds obsolete custom architecture mechanisms, and ln-646 validates physical project structure. Run ln-620-codebase-auditor to scan all 9 categories in parallel.
Audit skills in 5 groups: documentation quality (structure, semantics, fact-checking, inline code documentation), codebase health (security boundaries, delivery gates, DRY/KISS/YAGNI, maintainability hotspots, dependency/reuse risk, dead code, diagnosability, concurrency, lifecycle/config), test suites (business logic, E2E coverage, value scoring, coverage gaps, isolation, manual test quality, test structure), architecture (patterns, layer ownership, API contracts, dependency topology, modernization, project structure, configuration boundaries), and persistence performance (query efficiency, transactions, runtime, resource lifecycle).
</details> <details> <summary><b>How does it handle multi-stack or polyglot projects?</b></summary>Bootstrap skills (ln-7XX) support React, .NET, and Python project structures. Audit skills are language-aware — ln-622-build-delivery-gate-auditor checks compiler/type errors across stacks, ln-625-dependency-reuse-auditor scans npm, NuGet, and pip packages, and ln-651-query-efficiency-auditor catches N+1 queries regardless of ORM.
The active host model performs orchestration and verification. For code and story reviews, skills can delegate to the other CLI agent: Claude-hosted runs use Codex as advisor, while Codex-hosted runs use Claude as advisor.
</details> <details> <summary><b>What MCP servers are included?</b></summary>Four bundled MCP servers are published on npm: hex-line-mcp (hash-verified file editing), hex-graph-mcp (code knowledge graph), hex-ssh-mcp (hash-verified remote editing and SFTP over SSH), and hex-research-mcp (markdown research graph indexing). Skills work without MCP servers (fallback to built-in tools), but MCP improves edit accuracy and reduces token usage. See MCP Servers for setup.
Every line hex-line-mcp reads carries a content hash. When the agent edits, it must reference these hashes — proving it sees current file content, not a stale version from earlier in the conversation. This eliminates a common failure mode where AI agents overwrite recent changes with outdated context. Hooks auto-steer Claude from built-in tools to hex-line equivalents so the protection is always active.
Through the Orchestrator-Worker pattern plus persisted runtime state. Instead of feeding the entire codebase into one prompt, orchestrators advance stages from coordinator artifacts, coordinators consume worker artifacts, and workers execute with minimal, targeted context. Each layer loads only the files it needs and can resume from checkpoints instead of replaying long chat history.
</details> <details> <summary><b>How much context/tokens does it use?</b></summary>Skills are designed for token efficiency. Each worker loads only the files it needs via targeted context — no full-codebase prompts. Orchestrators advance stages from machine-readable artifacts instead of replaying chat history, and workers resume from checkpoints instead of reprocessing. See TOKEN_EFFICIENCY_PATTERNS.md for the patterns used.
</details> <details> <summary><b>Can I customize or create my own skills?</b></summary>Yes. Skills are SKILL.md files under plugins/<plugin>/skills/<skill>/. Single-skill support files live in that skill's references/; reusable shared resources live in root shared/ and are copied into exact skill-local targets listed in tools/marketplace/shared-registry.json with node tools/marketplace/shared.mjs sync. Validate with node tools/marketplace/validate.mjs. Reusable capabilities should use the plugin-first skill structure. See SKILL_ARCHITECTURE_GUIDE.md for the development workflow, shared-resource rules, 4-level hierarchy (L0 -> L1 -> L2 -> L3), and writing guidelines.
Yes. Claude and Codex have separate native marketplace surfaces in this repo. Install each agent independently through its own plugin flow; do not expose both surfaces to the same runtime at once. See AI Review Models > Sharing skills & MCP between agents for config paths.
</details>What's Inside
<details> <summary><b>Full Skill Tree</b></summary>claude-code-skills/ # MARKETPLACE
|-- plugins/ # INSTALLABLE PLUGINS + SELF-CONTAINED SKILLS
|-- shared/ # CANONICAL MULTI-SKILL DEVELOPMENT SOURCES
|-- tools/marketplace/shared-registry.json # shared -> skill-local references map
| |
| | ┌─ Plugin: agile-workflow ──────────────────────┐
| |
| |-- ln-2XX-*/ # PLANNING
| |-- ln-200-scope-decomposer/ # TOP: scope -> Epics -> Stories (one command)
| |-- ln-201-opportunity-discoverer/ # Traffic-First KILL funnel for growth direction
| |-- ln-210-epic-coordinator/ # CREATE/REPLAN 3-7 Epics
| |-- ln-220-story-coordinator/ # CREATE/REPLAN Stories + standards research
| | |-- ln-221-story-creator/ # Creates from IDEAL plan
| | |-- ln-222-story-replanner/ # Replans when requirements change
| |-- ln-230-story-prioritizer/ # RICE prioritization + market research
|
|-- ln-3XX-*/ # TASK MANAGEMENT
| |-- ln-300-task-coordinator/ # Artifact-first task planning coordinator
| | |-- ln-301-task-creator/ # Stateful task-plan worker (create)
| | |-- ln-302-task-replanner/ # Stateful task-plan worker (replan)
| |-- ln-310-multi-agent-validator/ # 20 criteria (8 groups), penalty points system + inline agent review
|
|-- ln-4XX-*/ # EXECUTION
| |-- ln-400-story-executor/ # Artifact-first execution coordinator
| |-- ln-401-task-executor/ # Stateful implementation worker
| |-- ln-402-task-reviewer/ # Stateful review worker and final task outcome
| |-- ln-403-task-rework/ # Stateful rework worker
| |-- ln-404-test-executor/ # Stateful test execution worker
|
|-- ln-5XX-*/ # QUALITY
| |-- ln-500-story-quality-gate/ # Thin orchestrator: verdict + Quality Score
| |-- ln-510-quality-coordinator/ # Artifact-first quality coordinator
| | |-- ln-511-code-quality-checker/ # Stateful quality worker
| | |-- ln-512-tech-debt-cleaner/ # Stateful autofix worker
| | |-- ln-513-regression-checker/ # Stateful regression worker
| | |-- ln-514-test-log-analyzer/ # Stateful log-analysis worker
| |-- ln-520-test-planner/ # Artifact-first test-planning coordinator
| | |-- ln-521-test-researcher/ # Stateful research worker
| | |-- ln-522-manual-tester/ # Stateful manual-testing worker
| | |-- ln-523-auto-test-planner/ # Stateful automated test-planning worker
|
|-- ln-10XX-*/ # ORCHESTRATION
| |-- ln-1000-pipeline-orchestrator/ # L0 Meta: coordinator artifacts drive the 4-stage pipeline
|
| └──────────────────────────────────────────────┘
| ┌─ Plugin: documentation-pipeline ──────────────┐
|
|-- ln-1XX-*/ # DOCUMENTATION
| |-- ln-100-documents-pipeline/ # L1 Orchestrator: complete docs in one command
| |-- ln-110-project-docs-coordinator/ # Detects project type, delegates to workers
| | |-- ln-111-root-docs-creator/ # AGENTS.md, CLAUDE.md, principles.md
| | |-- ln-112-project-core-creator/ # requirements.md, architecture.md
| | |-- ln-113-backend-docs-creator/ # api_spec.md, database_schema.md
| | |-- ln-114-frontend-docs-creator/ # design_guidelines.md
| | |-- ln-115-devops-docs-creator/ # infrastructure.md, runbook.md
| |-- ln-120-reference-docs-creator/ # ADRs, guides, manuals structure
| |-- ln-130-tasks-docs-creator/ # kanban_board.md, task provider setup
| |-- ln-140-test-docs-creator/ # testing-strategy.md
| |-- ln-160-docs-skill-extractor/ # Scan docs, classify, extract to .claude/commands
| | |-- ln-161-skill-creator/ # Transform doc sections into commands
| | |-- ln-162-skill-reviewer/ # Review SKILL.md and .claude/commands quality
|
| └──────────────────────────────────────────────┘
| ┌─ Plugin: codebase-audit-suite ────────────────┐
|
|-- ln-6XX-*/ # AUDIT
| |-- ln-610-docs-auditor/ # Documentation audit coordinator (4 workers)
| | |-- ln-611-docs-structure-auditor/ # Hierarchy, SSOT, compression, freshness
| | |-- ln-612-semantic-content-auditor/ # Scope alignment
| | |-- ln-613-code-comments-auditor/ # WHY-not-WHAT, density, docstrings
| | |-- ln-614-docs-fact-checker/ # Claims extraction, cross-doc verification
| |-- ln-620-codebase-auditor/ # 9 parallel auditors:
| | |-- ln-621-security-boundary-auditor/ # Security boundaries
| | |-- ln-622-build-delivery-gate-auditor/ # Build/delivery gates
| | |-- ln-623-duplication-overabstraction-auditor/# Duplication, over-abstraction
| | |-- ln-624-code-maintainability-hotspot-auditor/ # Maintainability hotspots
| | |-- ln-625-dependency-reuse-auditor/ # Dependency and reuse risk
| | |-- ln-626-dead-code-pruning-auditor/ # Dead code pruning
| | |-- ln-627-diagnosability-auditor/ # Diagnosability
| | |-- ln-628-concurrency-correctness-auditor/ # Concurrency correctness
| | |-- ln-629-runtime-lifecycle-config-auditor/ # Runtime lifecycle and config
| |-- ln-630-test-auditor/ # 7 test auditors:
| | |-- ln-631-test-business-logic-auditor/ # Framework vs business logic tests
| | |-- ln-632-test-e2e-priority-auditor/ # E2E coverage for critical paths
| | |-- ln-633-test-value-auditor/ # Risk-based test value scoring
| | |-- ln-634-test-coverage-auditor/ # Missing tests for critical paths
| | |-- ln-635-test-isolation-auditor/ # Isolation + anti-patterns
| | |-- ln-636-manual-test-auditor/ # Manual test quality (harness, golden files, fail-fast)
| | |-- ln-637-test-structure-auditor/ # Test file organization + directory layout
| |-- ln-640-pattern-evolution-auditor/ # Architectural pattern analysis + 4-score model
| | |-- ln-641-pattern-fitness-auditor/ # Pattern fitness worker
| | |-- ln-642-layer-ownership-boundary-auditor/# Layer and ownership boundaries
| | |-- ln-643-api-contract-auditor/ # Layer leakage, missing DTOs
| | |-- ln-644-dependency-topology-auditor/ # Topology cycles, coupling metrics (Ca/Ce/I)
| | |-- ln-645-architecture-modernization-auditor/ # Architecture modernization
| | |-- ln-646-project-structure-auditor/ # Physical architecture structure
| | |-- ln-647-configuration-boundary-auditor/ # Configuration boundary architecture
| |-- ln-650-persistence-performance-auditor/ # DB performance coordinator:
| | |-- ln-651-query-efficiency-auditor/ # N+1, over-fetching, missing bulk ops
| | |-- ln-652-transaction-correctness-auditor/ # Scope, rollback, long-held txns
| | |-- ln-653-runtime-performance-auditor/ # Blocking IO, allocations, sync sleep
| | |-- ln-654-resource-lifecycle-auditor/ # Session scope mismatch, pool config, cleanup
|
| └──────────────────────────────────────────────┘
| ┌─ Plugin: project-bootstrap ───────────────────┐
|
|-- ln-7XX-*/ # BOOTSTRAP
| |-- ln-700-project-bootstrap/ # L1: CREATE or TRANSFORM project
| |-- ln-720-structure-migrator/ # SCAFFOLD or RESTRUCTURE to Clean Architecture
| | |-- ln-721-frontend-restructure/ # React component-based architecture
| | |-- ln-722-backend-generator/ # .NET Clean Architecture from entities
| | |-- ln-723-seed-data-generator/ # Generate seed data from ORM schemas
| | |-- ln-724-artifact-cleaner/ # Remove platform-specific artifacts
| |-- ln-730-devops-setup/ # Docker, CI/CD, env
| | |-- ln-731-docker-generator/ # Dockerfiles, docker-compose
| | |-- ln-732-cicd-generator/ # GitHub Actions
| | |-- ln-733-env-configurator/ # .env.example
| |-- ln-740-quality-setup/ # Linters, pre-commit, tests
| | |-- ln-741-linter-configurator/ # ESLint, Prettier, Ruff, mypy
| | |-- ln-742-precommit-setup/ # Husky, lint-staged, commitlint
| | |-- ln-743-test-infrastructure/ # Vitest, xUnit, pytest setup
| |-- ln-760-security-setup/ # Security scanning
| | |-- ln-761-secret-scanner/ # Detect hardcoded secrets
| |-- ln-770-crosscutting-setup/ # Logging, CORS, health checks
| | |-- ln-771-logging-configurator/ # Structured JSON logging
| | |-- ln-772-error-handler-setup/ # Global exception middleware
| | |-- ln-773-cors-configurator/ # CORS policy config
| | |-- ln-774-healthcheck-setup/ # K8s readiness/liveness probes
| | |-- ln-775-api-docs-generator/ # Swagger/OpenAPI docs
| |-- ln-780-bootstrap-verifier/ # Build, test, Docker verification
| | |-- ln-781-build-verifier/ # Verify compilation
| | |-- ln-782-test-runner/ # Run test suites
| | |-- ln-783-container-launcher/ # Docker health check
|
| └──────────────────────────────────────────────┘
| ┌─ Plugin: optimization-suite ──────────────────┐
|
|-- ln-8XX-*/ # OPTIMIZATION
| |-- ln-810-performance-optimizer/ # Performance optimization:
| | |-- ln-811-performance-profiler/ # Full-stack request tracing, bottleneck classification
| | |-- ln-812-optimization-researcher/ # Competitive benchmarks, solution research, hypotheses
| | |-- ln-813-optimization-plan-validator/ # Agent-validated plan review (Claude/Codex)
| | |-- ln-814-optimization-executor/ # Strike-first hypothesis execution (keep/discard)
| |-- ln-820-dependency-optimization-coordinator/ # Dependency upgrades:
| | |-- ln-821-npm-upgrader/ # npm/yarn/pnpm with breaking change handling
| | |-- ln-822-nuget-upgrader/ # .NET NuGet with migration support
| | |-- ln-823-pip-upgrader/ # pip/poetry/pipenv with security audit
| |-- ln-830-code-modernization-coordinator/ # Code modernization:
| | |-- ln-831-oss-replacer/ # Replace custom code with OSS packages
| | |-- ln-832-bundle-optimizer/ # JS/TS bundle size reduction
| |-- ln-840-benchmark-compare/ # Canonical internal A/B: built-in vs hex-line, plus optional third Claude-compatible profile
|
| └──────────────────────────────────────────────┘
| ┌─ Plugin: community-engagement ───────────────┐
|
|-- ln-9XX-*/ # COMMUNITY ENGAGEMENT
| |-- ln-910-community-engagement/ # L2: Analyze health, consult strategy, delegate
| |-- ln-911-github-triager/ # Triage issues/PRs/discussions
| |-- ln-912-community-announcer/ # Compose + publish GitHub Discussion announcements
| |-- ln-913-community-debater/ # Launch RFC/debate/poll discussions
| |-- ln-914-community-responder/ # Respond to unanswered discussions/issues
|
| └──────────────────────────────────────────────┘
| ┌─ Plugin: setup-environment ────────────────────┐
|
|-- ln-001-push-all/ # Commit and push all changes in one command
|-- ln-002-session-analyzer/ # Analyze sessions for optimization opportunities
|-- ln-0XX-*/ # SETUP ENVIRONMENT
| |-- ln-010-dev-environment-setup/ # L2: Full environment setup coordinator
| |-- ln-011-agent-installer/ # Install/update Codex and Claude CLI
| |-- ln-012-mcp-configurator/ # Claude-side MCP setup: registration, hooks, permissions, migrations
| |-- ln-013-config-syncer/ # Align marketplace plugins, MCP state, and Codex defaults
| |-- ln-014-agent-instructions-manager/ # Single owner of CLAUDE.md/AGENTS.md creation and audit
| |-- ln-015-hex-line-uninstaller/ # Standalone cleanup for Claude-side hex-line integration
|-- ln-021-codegraph/ # Code knowledge graph for dependency analysis & impact checking
|-- ln-022-researchgraph/ # Research hypothesis/goal graph queries & audits
|
| └──────────────────────────────────────────────┘
|
|
|-- docs/
| |-- architecture/ # Skill patterns & delegation runtime
| |-- best-practice/ # Claude Code usage tips & component selection
| |-- standards/ # Documentation & README standards
|-- AGENTS.md # Canonical agent-facing repo map
|-- CLAUDE.md # Derived Anthropic entrypoint
Links
| Documentation | AGENTS.md |
| Architecture | SKILL_ARCHITECTURE_GUIDE.md |
| Agent Delegation | AGENT_DELEGATION_PLATFORM_GUIDE.md |
| Component Selection | COMPONENT_SELECTION.md |
| Workflow Tips | WORKFLOW_TIPS.md |
| Discussions | GitHub Discussions |
| Issues | GitHub Issues |
| Contributing | CONTRIBUTING.md |
| Browse Skills | skills.sh |
Papers, docs, and methodologies studied and implemented in the skill architecture.
| Source | Learned | Changed |
|---|---|---|
| STAR Framework (2025) | Forced goal articulation: +85pp accuracy; structured reasoning > context injection 2.83x | goal_articulation_gate.md — qualitative goal-articulation gate for relevant skills and templates |
| Building Effective Agents (Anthropic, 2024) | Orchestrator-Worker, prompt chaining, evaluator-optimizer patterns | Core 4-level hierarchy (L0→L3), single responsibility per skill |
| Multi-Agent Research System (Anthropic, 2025) | Production orchestration: 90.2% perf improvement with specialized agents | ln-1000 pipeline orchestrator, parallel agent reviews (ln-310, ln-510) |
| Scheduler Agent Supervisor (Microsoft) | Separation of scheduling, execution, and supervision | ln-400/ln-402/ln-500 executor-reviewer-gate split |
| DIATAXIS | 4-type docs: Tutorial / How-to / Reference / Explanation | Documentation levels in AGENTS.md/docs, progressive disclosure |
| Sinks, Not Pipes (Ian Bull, 2026) | "The architecture is the prompt" — AI agents can't reason about side-effect chains >2 levels deep; sinks (self-contained) > pipes (cascading) | ai_ready_architecture.md — cascade depth, architectural honesty, flat orchestration checks across relevant skills |
| Test Desiderata (Kent Beck, 2019) | 12 properties of valuable tests — behavioral, predictive, specific, inspiring, deterministic... No numerical targets, only usefulness | risk_based_testing_guide.md — 6 Test Usefulness Criteria (Risk Priority ≥15, Confidence ROI, Behavioral, Predictive, Specific, Non-Duplicative) |
| Vertical Slicing (Humanizing Work) | "Never split by architectural layer" | Foundation-First task ordering |
| Claude Code Picks (Amplifying AI, 2026) | Claude's tool preferences are learned maturity signals, not bias — Drizzle/Vitest/Zustand chosen for objective quality. Build-not-buy in 12/20 categories. "Correcting" valid preferences = recommending worse tools | Research-to-Action Gate in AGENTS.md — require concrete defect before turning research into skill changes |
| autoresearch (Karpathy, 2025) | Autoresearch loop: modify → benchmark → binary keep/discard; compound baselines; simplicity criterion (marginal gain + ugly code = discard) | ln-814-optimization-executor — keep/discard with adaptive thresholds, multi-file support, compound baselines, experiment log |
| The Complete Guide to Building Skills (Anthropic, 2026) | WHAT+WHEN descriptions, trigger testing, capability vs preference classification, negative triggers, 3-level progressive disclosure | Check #14 (trigger quality), negative trigger pattern, metadata.skill-type classification, functional DoD, M6 advisory |
License
Author: @levnikolaevich