• 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 adapters. Both surfaces are generated from the same canonical skills-catalog, but only one surface should be active inside a single runtime to avoid duplicate skill names.

MCP settings locations (for manual sharing):

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.

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.

Yes. MIT license, fully open source. No API keys, no paid services required. Claude Code CLI is the only prerequisite.

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.

# 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

Add the marketplace once, then install only what you need.

No. All skills work without Linear or any external tools. Linear integration is optional — when unavailable, skills fallback to a standalone flow using local markdown files (kanban_board.md) as the task management backend. No API keys, no 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 custom code that can be replaced by battle-tested open-source packages, and ln-646 validates project structure against framework-specific conventions. 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, build, DRY/KISS/YAGNI, complexity, dependencies, dead code, observability, concurrency, lifecycle), test suites (business logic, E2E coverage, value scoring, coverage gaps, isolation, manual test quality, test structure), architecture (patterns, layer boundaries, API contracts, dependency graphs, OSS replacements, project structure, env configuration), and persistence performance (query efficiency, transactions, runtime, resource lifecycle).

Bootstrap skills (ln-7XX) support React, .NET, and Python project structures. Audit skills are language-aware — ln-622-build-auditor checks compiler/type errors across stacks, ln-625-dependencies-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.

Three bundled MCP servers, all published on npm: hex-line-mcp (hash-verified file editing — every line carries a content hash, preventing stale-context corruption), hex-graph-mcp (code knowledge graph with symbol search, references, architecture analysis, and SCIP interop), and hex-ssh-mcp (hash-verified remote editing and SFTP over SSH). 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.

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.

Yes. Skills are SKILL.md files in skill directories. Legacy-compatible .claude/commands/*.md files still work as slash commands, but new reusable capabilities should use the skill structure. You can create standalone L3 workers or compose them into L2 coordinators and L1 orchestrators. See SKILL_ARCHITECTURE_GUIDE.md for the 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.

claude-code-skills/                      # MARKETPLACE
|-- skills-catalog/                              # ALL SKILLS + SHARED
|   |-- shared/                          # References, templates, agents
|   |
|   |  ┌─ 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-auditor/      # Secrets, SQL injection, XSS
|   |   |-- ln-622-build-auditor/         # Compiler/type errors
|   |   |-- ln-623-code-principles-auditor/# DRY/KISS/YAGNI, TODOs, DI
|   |   |-- ln-624-code-quality-auditor/  # Complexity, magic numbers
|   |   |-- ln-625-dependencies-auditor/  # Outdated packages + CVE vulnerabilities
|   |   |-- ln-626-dead-code-auditor/     # Unused code
|   |   |-- ln-627-observability-auditor/ # Logging, metrics
|   |   |-- ln-628-concurrency-auditor/   # Race conditions
|   |   |-- ln-629-lifecycle-auditor/     # Bootstrap, shutdown
|   |-- 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-analyzer/      # Pattern scoring worker
|   |   |-- ln-642-layer-boundary-auditor/# Layer violations, I/O isolation
|   |   |-- ln-643-api-contract-auditor/  # Layer leakage, missing DTOs
|   |   |-- ln-644-dependency-graph-auditor/ # Cycles, coupling metrics (Ca/Ce/I)
|   |   |-- ln-645-open-source-replacer/ # Goal-based OSS replacement audit + migration plan
|   |   |-- ln-646-project-structure-auditor/ # Physical structure audit with framework-specific rules
|   |   |-- ln-647-env-config-auditor/ # Env var config, sync, naming, startup validation
|   |-- 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-020-codegraph/                  # Code knowledge graph for dependency analysis & impact checking
|
|  └──────────────────────────────────────────────┘
|
|
|-- 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

Papers, docs, and methodologies studied and implemented in the skill architecture.