Agent-Skills.md

Agent Skills: moai

>

UncategorizedID: modu-ai/moai-adk/moai

Repository

modu-ai/moai-adk
modu-aiLicense: GPL-3.0
873151

Install this agent skill to your local

pnpm dlx add-skill https://github.com/modu-ai/moai-adk/tree/HEAD/.claude/skills/moai

Skill Files

Browse the full folder contents for moai.

Download Skill

Loading file tree…

.claude/skills/moai/SKILL.md

Skill Metadata

Name
moai
Description
>

Pre-execution Context

!git status --porcelain 2>/dev/null || true !git branch --show-current 2>/dev/null || true

Essential Files

.moai/config/config.yaml

Authority References

Rules and constraints governing all workflows are always loaded from these sources. Do NOT duplicate their content here:

  • Core identity, orchestration principles, agent catalog: CLAUDE.md
  • Quality gates, security boundaries: .claude/rules/moai/core/moai-constitution.md
  • SPEC workflow phases, token budgets: .claude/rules/moai/workflow/spec-workflow.md
  • Development methodologies (DDD/TDD): .claude/rules/moai/workflow/spec-workflow.md (Run Phase section)
  • Agent definitions: See CLAUDE.md Section 4. For agent creation, use builder-harness subagent (artifact_type=agent).
  • @MX tag rules and protocol: .claude/rules/moai/workflow/mx-tag-protocol.md

Intent Router

Raw User Input

$ARGUMENTS

Routing Instructions

[HARD] Route the Raw User Input above using the strict priority order below. Extract the FIRST WORD of the input for subcommand matching. All text after the subcommand keyword is CONTEXT to be passed to the matched workflow — it is NOT a routing signal and MUST NOT influence which workflow is selected.

Execution Mode Flags (mutually exclusive)

  • --team: Force Agent Teams mode for parallel execution
  • --solo: Force sub-agent mode (single agent per phase)
  • No flag: System auto-selects based on complexity thresholds (domains >= 3, files >= 10, or complexity score >= 7)

When no flag is provided, the system evaluates task complexity and automatically selects between team mode (for complex, multi-domain tasks) and sub-agent mode (for focused, single-domain tasks).

Priority 1: Explicit Subcommand Matching

[HARD] Extract the FIRST WORD from the Raw User Input section above. If it matches any subcommand below (or its alias), route to that workflow IMMEDIATELY. Do NOT analyze the remaining text for routing — it is context for the matched workflow:

  • brain (aliases: ideate, idea): Pre-spec ideation workflow — 7-phase idea-to-proposal pipeline with Claude Design handoff package. Runs BEFORE project and plan.
  • plan (aliases: spec): SPEC document creation workflow
  • run (aliases: impl): DDD/TDD implementation workflow (per quality.yaml development_mode)
  • sync (aliases: docs, pr): Documentation synchronization and PR creation
  • design (aliases: brief, brand): Hybrid design workflow (Claude Design import path A or code-based skill path B)
  • project (aliases: init): Project documentation generation
  • feedback (aliases: fb, bug, issue): GitHub issue creation
  • fix: Auto-fix errors in a single pass
  • loop: Iterative auto-fix until completion marker detected
  • mx: MX tag scan and annotation for codebase
  • review (aliases: code-review): Code review with security and MX tag compliance
  • clean (aliases: dead-code): Identify and safely remove dead code
  • codemaps: Generate architecture documentation in .moai/project/codemaps/
  • coverage (aliases: cov): Analyze test coverage and generate missing tests
  • e2e (aliases: e2e-test): Create and run E2E tests
  • gate (aliases: check, pre-commit): Lightweight pre-commit quality gate (lint+format+type-check+test)
  • security (aliases: audit, sec): Dedicated OWASP security audit with dependency scanning
  • harness (aliases: hrn, learn): V3R4 self-evolving harness lifecycle (status / apply / rollback <date> / disable) — slash-command-only surface; CLI verb path retired per SPEC-V3R4-HARNESS-001 (BC-V3R4-HARNESS-001-CLI-RETIREMENT)
  • release-update (aliases: cc-update, release-track) (dev-only): CC upstream change tracker → update plan + docs-site 4-locale sync

Priority 2: SPEC-ID Detection

Only if Priority 1 did not match: Check if the Raw User Input contains a pattern matching SPEC-XXX (such as SPEC-AUTH-001). If found, route to the run workflow automatically. The SPEC-ID becomes the target for DDD/TDD implementation.

Priority 3: Natural Language Classification

Only if BOTH Priority 1 AND Priority 2 did not match: Classify the intent of the ENTIRE Raw User Input as natural language. This priority is NEVER reached when the first word matches a known subcommand.

  • Planning and design language (design, architect, plan, spec, requirements, feature request) routes to plan
  • Quality gate language (lint, format, check, pre-commit, quality gate) routes to gate
  • Security language (security, audit, owasp, vulnerability, injection, xss, csrf) routes to security
  • Error and fix language (fix, error, bug, broken, failing, lint) routes to fix
  • Iterative and repeat language (keep fixing, until done, repeat, iterate, all errors) routes to loop
  • Documentation language (document, sync, docs, readme, changelog, PR) routes to sync or project
  • Feedback and bug report language (report, feedback, suggestion, issue) routes to feedback
  • MX tag language (mx tag, annotation, code context, legacy annotate) routes to mx
  • Implementation language (implement, build, create, add, develop) with clear scope routes to moai (default autonomous)

Priority 4: Default Behavior

If the intent remains ambiguous after all priority checks, use AskUserQuestion to present the top 2-3 matching workflows and let the user choose.

If the intent is clearly a development task with no specific routing signal, default to the moai workflow (plan -> run -> sync pipeline) for full autonomous execution.

Workflow Quick Reference

plan - SPEC Document Creation

Purpose: Create comprehensive specification documents using EARS format with Research-Plan-Annotate cycle. Phases: Deep Research (research.md) -> SPEC Planning -> Annotation Cycle (1-6 iterations) -> SPEC Creation Agents: manager-spec (primary), Explore (research), manager-git (conditional) Flags: --worktree, --branch, --resume SPEC-XXX, --team, --issue (opt-in; default skips GitHub Issue creation per SPEC-V3R5-LATE-BRANCH-001 REQ-LB-009) For detailed orchestration: Read ${CLAUDE_SKILL_DIR}/workflows/plan.md (team mode: ${CLAUDE_SKILL_DIR}/team/plan.md)

run - DDD/TDD Implementation

Purpose: Implement SPEC requirements through configured development methodology. Agents: manager-strategy, manager-develop (cycle_type=ddd|tdd per quality.yaml), manager-quality, manager-git Flags: --resume SPEC-XXX, --team For detailed orchestration: Read ${CLAUDE_SKILL_DIR}/workflows/run.md (team mode: ${CLAUDE_SKILL_DIR}/team/run.md)

sync - Documentation Sync and PR

Purpose: Synchronize documentation with code changes and prepare pull requests. Agents: manager-docs (primary), manager-quality, manager-git Modes: auto, force, status, project. Flags: --merge, --skip-mx For detailed orchestration: Read ${CLAUDE_SKILL_DIR}/workflows/sync.md (team mode: ${CLAUDE_SKILL_DIR}/team/sync.md)

gate - Pre-Commit Quality Gate

Purpose: Lightweight pre-commit quality check running lint, format, type-check, and tests in parallel. Also integrated into run (Phase 2.75) and sync (Phase 0) workflows as automatic pre-checks. Agents: Direct execution (no agent delegation) Flags: --fix, --staged, --file PATH Integration: Automatically invoked by run workflow (Phase 2.75) and sync workflow (Phase 0.0.1) with --fix behavior. For detailed orchestration: Read ${CLAUDE_SKILL_DIR}/workflows/gate.md

security - OWASP Security Audit

Purpose: Dedicated security audit with OWASP Top 10 analysis, dependency scanning, secrets detection, and data isolation checks. Agents: expert-security (primary) Flags: --full, --deps, --secrets, --file PATH, --branch BRANCH For detailed orchestration: Read ${CLAUDE_SKILL_DIR}/workflows/security.md

fix - Auto-Fix Errors

Purpose: Autonomously detect and fix LSP errors, linting issues, and type errors. Agents: manager-quality (diagnostic-mode), expert-backend/expert-frontend (fixes) Flags: --dry, --sequential, --level N, --resume, --team For detailed orchestration: Read ${CLAUDE_SKILL_DIR}/workflows/fix.md

loop - Iterative Auto-Fix

Purpose: Repeatedly fix issues until completion marker detected or max iterations reached. Agents: manager-quality (diagnostic-mode), expert-backend, expert-frontend, manager-develop Flags: --max N, --auto-fix, --seq For detailed orchestration: Read ${CLAUDE_SKILL_DIR}/workflows/loop.md

mx - MX Tag Scan and Annotation

Purpose: Scan codebase and add @MX code-level annotations for AI agent context. Agents: Explore (scan), expert-backend (annotation) Flags: --all, --dry, --priority P1-P4, --force, --team For detailed orchestration: Read ${CLAUDE_SKILL_DIR}/workflows/mx.md

review - Code Review

Purpose: Multi-perspective code review with security, performance, quality, and UX analysis. Agents: manager-quality (primary), expert-security Flags: --staged, --branch, --security, --team For detailed orchestration: Read ${CLAUDE_SKILL_DIR}/workflows/review.md (team mode: ${CLAUDE_SKILL_DIR}/team/review.md)

clean - Dead Code Removal

Purpose: Identify and safely remove unused code with test verification. Agents: expert-refactoring, manager-develop Flags: --dry, --safe-only, --file PATH For detailed orchestration: Read ${CLAUDE_SKILL_DIR}/workflows/clean.md

codemaps - Architecture Documentation

Purpose: Scan codebase and generate architecture documentation. Agents: Explore, manager-docs Flags: --force, --area AREA For detailed orchestration: Read ${CLAUDE_SKILL_DIR}/workflows/codemaps.md

coverage - Test Coverage Analysis

Purpose: Analyze test coverage gaps and generate missing tests. Agents: manager-develop (cycle_type=tdd) Flags: --target N, --file PATH, --report For detailed orchestration: Read ${CLAUDE_SKILL_DIR}/workflows/coverage.md

e2e - End-to-End Testing

Purpose: Create and run E2E tests using Chrome, Playwright, or Agent Browser. Agents: manager-develop (cycle_type=tdd), expert-frontend Flags: --record, --url URL, --journey NAME For detailed orchestration: Read ${CLAUDE_SKILL_DIR}/workflows/e2e.md

design - Hybrid Design Workflow

Purpose: Produce web/brand design artifacts via Claude Design import (path A) or code-based skill pipeline (path B). Integrates brand context from .moai/project/brand/ and design briefs from .moai/design/. Agents: manager-spec (BRIEF), expert-frontend (implementation), evaluator-active (GAN loop scoring) Skills: moai-domain-copywriting, moai-domain-brand-design, moai-workflow-design, moai-workflow-gan-loop Flags: --path A|B, --harness thorough|standard, --brief BRIEF-XXX For detailed orchestration: Read ${CLAUDE_SKILL_DIR}/workflows/design.md

(default) - MoAI Autonomous Workflow

Purpose: Full autonomous research -> plan -> annotate -> run -> sync pipeline. Phases: Parallel Exploration (research.md) -> SPEC Generation -> Annotation Cycle -> Implementation -> Sync Agents: Explore, manager-spec, manager-develop, manager-quality, manager-docs, manager-git Flags: --loop, --max N, --branch, --pr, --resume SPEC-XXX, --team, --solo, --issue (opt-in; default skips GitHub Issue creation per SPEC-V3R5-LATE-BRANCH-001 REQ-LB-009) For detailed orchestration: Read ${CLAUDE_SKILL_DIR}/workflows/moai.md

project - Project Documentation

Purpose: Generate project documentation by analyzing the existing codebase. Agents: Explore, manager-docs, expert-devops (optional) Output: product.md, structure.md, tech.md in .moai/project/ For detailed orchestration: Read ${CLAUDE_SKILL_DIR}/workflows/project.md

feedback - GitHub Issue Creation

Purpose: Collect user feedback and create GitHub issues. Agents: manager-quality For detailed orchestration: Read ${CLAUDE_SKILL_DIR}/workflows/feedback.md

harness - V3R4 Self-Evolving Harness Lifecycle

Purpose: Surface the harness learning subsystem (observer, 4-tier proposal ladder, 5-layer safety pipeline) to the user via the slash command path. Owns all lifecycle verbs (status / apply / rollback / disable) entirely within the workflow body using file-system operations — no Go binary subcommand invoked. Tier-4 application is gated by orchestrator-issued AskUserQuestion per REQ-HRN-FND-004. Skills: moai-harness-learner (Tier-4 surfacing companion), moai-meta-harness (project-specific harness generation, indirect) Verbs: status (tier distribution + telemetry) | apply (next Tier-4 proposal → AskUserQuestion → 5-layer pipeline → snapshot + write) | rollback <YYYY-MM-DD> (restore snapshot) | disable (set learning.enabled: false) Artifacts: .moai/harness/usage-log.jsonl, .moai/harness/proposals/, .moai/harness/learning-history/snapshots/, .moai/harness/learning-history/applied/, .moai/harness/learning-history/frozen-guard-violations.jsonl Authoritative SPEC: SPEC-V3R4-HARNESS-001 (supersedes V3R3-HARNESS-001, V3R3-HARNESS-LEARNING-001, V3R3-PROJECT-HARNESS-001) For detailed orchestration: Read ${CLAUDE_SKILL_DIR}/workflows/harness.md

release-update - CC Upstream Change Tracker (dev-only)

Purpose: Track Claude Code release notes since last analyzed version, classify by impact tier, generate update plan, sync docs-site 4-locale + README, open PR. Agents: manager-docs (Phase 6 docs sync), manager-git (Phase 7 PR) Flags: --since vX.Y.Z, --dry, --report-only, --docs-only, --master-spec State: .moai/state/last-cc-version.json For detailed orchestration: Read ${CLAUDE_SKILL_DIR}/workflows/release-update.md NOT distributed to user projects (dev-only; entry: .claude/commands/97-release-update.md)

Execution Directive

When this skill is activated, execute the following steps in order:

Step 1 - Parse Arguments: Extract subcommand keywords and flags from the Raw User Input. Recognized global flags: --resume [ID], --seq, --team, --solo. Also detect ultrathink keyword in the input text.

CRITICAL: Deep analysis mode:

  • ultrathink keyword detected → Activate Claude's native extended reasoning (high effort mode). This is native Claude behavior with no MCP dependency.

Step 1.5 - Flag-Subcommand Compatibility Validation: [HARD] After parsing the subcommand and flags (Step 1), validate flag-subcommand compatibility BEFORE routing. If a forbidden combination is detected, STOP all further processing and output an error in the user's conversation_language. Do NOT proceed to Step 2.

Forbidden flag-subcommand combinations:

| Flag | Allowed subcommands | Forbidden subcommands | |------|---------------------|------------------------| | --worktree | plan, default (autonomous) | run, sync | | --branch | plan, default (autonomous) | run, sync |

Rationale: --worktree (and --branch) provision the workspace at SPEC initialization. /moai plan is the sole entry point that creates the worktree/branch. /moai run and /moai sync MUST operate within the worktree/branch already established during plan. Re-creating during run/sync corrupts the SPEC lifecycle and is rejected at the router level.

Error message template (Korean conversation_language; substitute the actual flag and subcommand):

에러: --worktree 플래그는 /moai plan 전용입니다.
/moai run 과 /moai sync 는 plan 단계에서 생성된 기존 worktree/branch를 재사용합니다.

올바른 사용법:
  /moai plan SPEC-XXX --worktree    (worktree 생성)
  /moai run SPEC-XXX                (기존 worktree/branch 재사용)
  /moai sync SPEC-XXX               (기존 worktree/branch 재사용)

다시 실행하려면 --worktree 플래그를 제거한 형태로 호출하세요.

For English (en conversation_language), translate the message; the structure remains identical.

Step 2 - Route to Workflow: Apply the Intent Router (Priority 1 through Priority 4) to determine the target workflow. If ambiguous, use AskUserQuestion to clarify with the user.

Step 2.5 - Project Documentation Check: Before executing plan, run, sync, fix, loop, or default workflows, verify project documentation exists by checking for .moai/project/product.md. If product.md does NOT exist, use AskUserQuestion to ask the user (in their conversation_language):

Question: Project documentation not found. Would you like to create it first? Options:

  • Create project documentation (Recommended): Generates product.md, structure.md, tech.md through a guided interview. This helps MoAI understand your project context for better results in all subsequent workflows.
  • Skip and continue: Proceed without project documentation. MoAI will have less context about your project.

This check does NOT apply to: project, feedback subcommands.

[HARD] Beginner-Friendly Option Design: All AskUserQuestion calls throughout MoAI workflows MUST follow these rules:

  • The first option MUST always be the recommended choice, clearly marked with "(Recommended)" suffix
  • Every option MUST include a detailed description explaining what it does and its implications

Step 3 - Load Workflow Details: If --team flag was parsed AND ${CLAUDE_SKILL_DIR}/team/<name>.md exists for the target subcommand, read the team workflow file instead of the solo workflow. Otherwise read workflows/<name>.md. The Quick Reference section above shows both paths for each subcommand that supports team mode.

Step 4 - Read Configuration: Load relevant configuration from .moai/config/config.yaml and section files as needed.

Step 5 - Initialize Task Tracking: Use TaskCreate to register discovered work items with pending status.

Step 6 - Execute Workflow Phases: Follow the workflow-specific phase instructions. Delegate all implementation to appropriate agents via Agent(). Collect user approvals at designated checkpoints via AskUserQuestion.

Step 7 - Track Progress: Update task status using TaskUpdate as work progresses (pending to in_progress to completed).

Step 8 - Present Results: Display results to the user in their conversation_language using Markdown format.

Step 9 - Add Completion Marker: When all workflow phases complete successfully, add the appropriate completion marker (<moai>DONE</moai> or <moai>COMPLETE</moai>).

Step 10 - Guide Next Steps: Use AskUserQuestion to present the user with logical next actions based on the completed workflow.

Version: 2.6.0 Last Updated: 2026-02-25

<!-- moai:evolvable-start id="rationalizations" -->

Common Rationalizations

| Rationalization | Reality | |---|---| | "I will just implement this directly, delegation is overhead" | MoAI is an orchestrator. Direct implementation bypasses the quality gates agents enforce. | | "The user's intent is obvious, no need for a Socratic interview" | Ambiguous verbs (clean, fix, improve) almost always produce wrong scope. Rule 5 exists because obvious is often wrong. | | "This is a small change, Approach-First is unnecessary" | Small changes still touch files the user cares about. One sentence of approach costs nothing and prevents rework. | | "I can run /moai run without a SPEC, it is just a tweak" | Without a SPEC, there is no acceptance criterion to check. Every run without a SPEC silently degrades quality tracking. | | "Parallel agents will just race, sequential is safer" | Independent tool calls are explicitly required to run in parallel. Sequentializing them wastes user time. | | "I will respond in English since it is technical" | Conversation language is a HARD rule. User-facing output must match the configured language, always. | | "Schema 로드가 귀찮으니 이번엔 산문으로 질문하자" | AskUserQuestion/Task* 는 deferred tool. ToolSearch 한 번으로 session 전체 사용 가능. 산문 질문은 HARD 위반 (CLAUDE.md §1, §8 Deferred Tool Preload Protocol). | | "짧은 확인 질문은 산문으로 처리해도 된다" | 모든 user-facing 질문은 AskUserQuestion 경유 강제. "짧은 질문"은 예외 아님. Self-check: 응답에 "?" 있으면 AskUserQuestion 호출 동반 필수. |

<!-- moai:evolvable-end --> <!-- moai:evolvable-start id="red-flags" -->

Red Flags

  • MoAI writes code directly instead of delegating to a specialized agent
  • Response in English when conversation_language is not English
  • Multiple independent tool calls executed sequentially in separate messages
  • AskUserQuestion with more than 4 options or containing emoji
  • Agent invocation prompt contains absolute paths to the main project when isolation is worktree
  • /moai run executed without a corresponding SPEC-XXX document
  • Response ends with "?" but no AskUserQuestion tool call accompanies it
  • Options listed as markdown (- A:, - B:, Option X:) without structured AskUserQuestion
  • Prose decision requests ("진행할까요?", "어느 것 선호?", "A or B?") instead of AskUserQuestion
  • First AskUserQuestion call in session without prior ToolSearch preload (produces InputValidationError)
  • Waiting for user's next message after prose question without AskUserQuestion tool call
<!-- moai:evolvable-end --> <!-- moai:evolvable-start id="verification" -->

Verification

  • [ ] User-facing response language matches conversation_language from language.yaml
  • [ ] Every independent tool call was launched in parallel (one message, multiple tool blocks)
  • [ ] Agent selection trace documents why this agent, not another, was chosen
  • [ ] No XML tags visible in user-facing output
  • [ ] For non-trivial tasks, approach was explained and approved before code changes
  • [ ] SPEC-ID is referenced when /moai run, /moai sync, or /moai fix is invoked
  • [ ] TodoList used to decompose multi-file changes (3+ files)
  • [ ] Session opened with ToolSearch preload of deferred tools (AskUserQuestion, TaskCreate/Update/List/Get)
  • [ ] Every response containing "?" is accompanied by a structured AskUserQuestion tool call
  • [ ] Option lists (- A:, - B:) are routed through AskUserQuestion, not markdown-only
  • [ ] No silent "wait for user input" state after prose question (§8 Deferred Tool Preload Protocol)
<!-- moai:evolvable-end -->