Instructions file
Schema and harness document for the CMDS LLM Wiki vault. Defines the 3-layer architecture (Raw Sources / Wiki / Schema), ingest-query-lint operations, file conventions, and frontmatter standards. This is the single source of truth for LLM behavior in this vault.
Skill Claude CodeCodex
Self-assess your agentic knowledge management system against the AKM Index v1.2 (scoring anchors unchanged since v1.1) — a universal 5-pillar × 25-criterion rubric (Prompt / Context / Harness / Loop engineering + Interop & Governance), scored 0-100 with behaviorally anchored levels and mandatory artifact evidence.…
Plugin Claude Code
Claude Code plugins by CMDSPACE (커맨드스페이스) - Research Pipeline, and more.
Plugin Claude Code
Autonomous Development Orchestrator with 3-Tier agent system (Opus/Sonnet/Haiku). Provides 13 specialized agents and 12 skills for the full SDLC — from planning and architecture through implementation, testing, review, and documentation.
Agent
Use this agent for system and code architecture design, technology decisions, and structural analysis. The architect evaluates trade-offs, selects patterns, designs data models, and produces architecture decision records (ADRs). Best for greenfield design, major refactors, and technology selection. Context: User needs…
Agent
Use this agent to automatically diagnose and fix build errors, compilation failures, type errors, and dependency issues. The build-fixer reads error logs, traces root causes, and applies targeted fixes. It supports iterative fixing — retrying until the build succeeds. Context: User has build errors they can't resolve…
Agent
Use this agent for thorough code quality review. Examines code for readability, maintainability, performance, naming conventions, and best practices. Produces structured review comments with severity levels. Unlike security-reviewer (which focuses on vulnerabilities), code-reviewer focuses on code quality and…
Agent
Use this agent as a Devil's Advocate to critically review plans, designs, code, and decisions. The critic finds logical flaws, hidden assumptions, edge cases, scalability issues, and failure modes that others miss. Unlike code-reviewer (which checks style and quality), the critic challenges fundamental assumptions and…
Agent
Use this agent for UI/UX component design and frontend architecture. The designer creates component hierarchies, handles layout and styling, ensures accessibility (a11y), and applies design system tokens. Works with React, Vue, Svelte, SwiftUI, and CSS frameworks (Tailwind, CSS-in-JS). Context: User needs a UI…
Agent
Use this agent for actual code implementation tasks. The executor writes production-quality code following existing project conventions, patterns, and style. It is the most frequently called agent — the "hands" of the system. Best for implementing features, fixing bugs, creating APIs, and writing business logic.…
Agent
Use this agent for fast codebase exploration, file discovery, dependency mapping, and structural analysis. The explorer quickly indexes project structure, finds relevant files, traces import chains, and builds a map of how code is organized. Optimized for speed — uses Haiku tier for rapid responses. Context: User…
Agent
Use this agent as the Master Orchestrator for complex, multi-step development tasks that require strategic planning and coordination. The planner analyzes requirements, decomposes work into subtasks, assigns appropriate tiers (Opus/Sonnet/Haiku), identifies dependencies and parallelization opportunities, and produces…
Agent
Use this agent for quality assurance testing — end-to-end tests, integration tests, regression tests, and edge case discovery. The QA tester thinks like a user and an adversary simultaneously, finding scenarios that break the system. Context: User wants E2E or integration tests user: "이 기능의 통합 테스트를 작성해줘" assistant…
Agent
Use this agent for technical research — evaluating libraries, comparing frameworks, investigating best practices, and gathering information from documentation and the web. The researcher produces evidence-based reports with recommendations, not opinions. Context: User needs to evaluate technology options user: "ORM으로…
Agent
Use this agent for security vulnerability analysis and security-focused code review. Examines code for OWASP Top 10 vulnerabilities, authentication/authorization flaws, injection risks, data exposure, and dependency vulnerabilities. Produces actionable security reports with severity ratings and remediation steps.…
Agent
Use this agent for test-driven development guidance. The TDD guide writes test cases FIRST, then guides implementation to satisfy them. It enforces the Red-Green-Refactor cycle, generates comprehensive test scenarios including edge cases, and tracks coverage. Context: User wants to implement a feature with TDD user…
Agent
Use this agent for documentation writing — README files, API documentation, code comments, architecture guides, changelogs, and inline documentation. The writer produces clear, concise, well-structured documentation that matches the project's existing docs style. Fast processing optimized for documentation output.…
Skill Claude CodeCodex
Fully autonomous execution mode. When triggered by '/autopilot', 'autopilot mode', '자율 실행', or '알아서 해줘', this skill takes a development task and autonomously plans, implements, tests, and delivers without user intervention. Uses the planner agent to decompose work, then dispatches appropriate agents in parallel where…
Skill Claude CodeCodex
Deep code review skill. When triggered by '/code-review', '코드 리뷰', 'review this code', or '코드 품질 점검', this skill performs a thorough code quality review with structured findings, severity ratings, and actionable recommendations.
Skill Claude CodeCodex
Task decomposition skill. When triggered by '/decompose', '작업 분해', 'break down this task', or '서브태스크로 나눠줘', this skill takes a complex task and recursively decomposes it into atomic, testable subtasks with tier assignments and parallelization opportunities.
Skill Claude CodeCodex
Hierarchical documentation generation skill. When triggered by '/deepinit', '문서 생성', 'generate docs', or '프로젝트 문서화', this skill analyzes the codebase and generates comprehensive documentation — README, architecture docs, API docs, and Mermaid diagrams.
Skill Claude CodeCodex
Multi-strategy deep code search skill. When triggered by '/deepsearch', '깊은 검색', 'deep search', or '코드 찾아줘', this skill applies multiple search strategies — structural (AST-like pattern), textual (regex), semantic (by meaning), and historical (git history) — to find code that simple grep can't.
Skill Claude CodeCodex
Task estimation skill. When triggered by '/estimate', '추정해줘', 'estimate effort', or '이거 얼마나 복잡해?', this skill analyzes a development task and estimates complexity, agent tier distribution, and provides a confidence level for the estimate.
Skill Claude CodeCodex
Deep focus mode using a single high-tier agent for complex problems requiring sustained reasoning. When triggered by '/focus', 'focus mode', '집중 모드', or '깊이 생각해줘', this skill dedicates a single Opus-tier agent to one problem without context switching, multi-agent coordination, or parallelization.