"architect-skill"
"Senior Software Architect & System Designer: Deterministic Tier 1/2 wizard, PRD/BRD/SRD requirements, Clean Architecture, distributed system design, senior frontend aesthetics, and zero-egress local-first security."
Safety scan
A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.
Badge for your README
Markdown
[](/listings/fiky0911--skill-architecture)HTML
<a href="/listings/fiky0911--skill-architecture"><img src="/api/listings/fiky0911--skill-architecture/badge.svg" alt="findsafeskills safety rating" /></a>Install
Install instructions from the README
As written by the repo's author. findsafeskills has not verified these steps.
The toolkit includes a zero-dependency installer and configuration engine compatible with Linux, macOS, and Windows across all 7 supported AI agents.
4.1 Zero-Clone Installation via npm & npx
The package is officially published on the NPM Registry (`npm i skill-architecture`). Users do not need to clone this repository. You can install or run `skill-architecture` across all AI agents with a single command:
Standard CLI Execution via `npx` (No Install Required):
# Install globally to all supported AI agents in one command npx skill-architecture --global # Target a specific AI agent directly (e.g., Antigravity, Cursor, or Claude) npx skill-architecture --global --agent antigravity npx skill-architecture --global --agent cursor npx skill-architecture --global --agent claude # Local workspace installation npx skill-architecture --local --target all
Install via `npm` (Global or Project Dependency):
# Install globally on your machine npm install -g skill-architecture skill-architecture --global # Or install as a dev dependency in your project npm install -D skill-architecture # or simply: npm i skill-architecture
Industry Standard via `npx skills` (skills.sh ecosystem by Vercel Labs):
Because this repository conforms strictly to the open `SKILL.md` specification, users of the universal agent skill manager can also install it natively:
# Add to all supported agents globally npx skills add skill-architecture -g # or from GitHub: npx skills add FIKY0911/skill-architecture -g # Add to a specific agent (e.g., Cursor or Claude Code) npx skills add skill-architecture --agent cursor npx skills add skill-architecture --agent claude-code
---
4.2 Direct ZIP Package Upload (Zero-Terminal / Web AI UI)
For users who want to use this skill directly inside web-based AI interfaces without using a terminal or command line:
1. **Download the standalone ZIP package**: - Download [`architect-skill.zip`](file:///home/fiky/Documents/Latihan Agent/SKILL/architect-skill/architect-skill.zip) directly from the repository / GitHub Releases or build it locally using `npm run bundle:zip`. 2. **Upload to Your AI Agent Environment**: - **Claude Desktop / Claude.ai Projects**: Attach or upload `architect-skill.zip` to your project knowledge base or custom instructions. - **Google Antigravity / Gemini Workspace**: Place the unzipped contents into `~/.gemini/skills/architect-skill` or upload via the skills manager. - **ChatGPT Custom GPT / Assistants**: Attach `architect-skill.zip` into the GPT's **Knowledge / Code Interpreter** storage. - **Cursor / VS Code IDE**: Unzip directly into your project's root folder.
The ZIP bundle contains the complete offline architecture rulebook, deterministic state machine, zero-egress safeguards, and industry reference standards in a compact ~450 KB archive.
---
4.3 Local & Monorepo Installation (For Contributors)
# Preview installation plan without touching disk (--dry-run) node scripts/install.js --global --dry-run # Install globally for all 7 AI agents node scripts/install.js --global # Install for a single specific AI agent node scripts/install.js --global --target antigravity node scripts/install.js --local --target cursor # Complete clean uninstallation node scripts/install.js --global --uninstall node scripts/install.js --local --uninstall
---
4.4 Recommended MCP Servers (Model Context Protocol)
To unlock the full potential of `skill-architecture` (especially for pulling official library documentation without guessing and generating zero-slop UI components), install the following recommended MCP servers into your agent's configuration (`mcp_config.json`):
1. Core Documentation & Verification MCPs
What was scanned
The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.
SKILL.md · Manifest · 328 lines
- ---
- name: "architect-skill"
- version: "1.1.0"
- description: "Senior Software Architect & System Designer: Deterministic Tier 1/2 wizard, PRD/BRD/SRD requirements, Clean Architecture, distributed system design, senior frontend aesthetics, and zero-egress local-first security."
- author: "John McCarthy & Architect Skill Contributors"
- license: "MIT"
- ---
- # ARCHITECT-SKILL: SYSTEM ORCHESTRATION INSTRUCTIONS
- ## Token‑Usage Guidelines
- - **max_input_tokens**: 150 – total tokens from the user prompt should not exceed this limit.
- - **max_output_tokens**: 30 – the skill should keep its responses concise.
- - **step_token_budget**: 30 – each internal step or sub‑prompt must stay within this budget.
- Keep prompts terse and avoid unnecessary context. Use references to existing documentation instead of repeating large texts.
- You are a Senior Software Architect and Foundational Systems Logician. You operate under the principles of **constructive minimalism**, **formal traceability**, and **zero assumptions**.
- When this skill is invoked, you guide the user through software architecture generation, requirements engineering, distributed system design, and senior frontend aesthetic curation, emitting an authoritative documentation bundle into `./docs/`.
- ---
- ## 🔒 Foundational Axioms & Security Constraints
- 1. **Zero-Egress Mandate**: You operate strictly within the local workspace. You must never initiate external network calls, transmit telemetry, or depend on remote APIs.
- 2. **Absolute Credential Protection**: You are strictly forbidden from inspecting, requesting, reading, or processing:
- - Environment files: `.env`, `.env.*`, `*.env`
- - Private keys and certificates: `*.pem`, `*.key`, `*.keystore`, `*.crt`, `*.p12`, `*.pfx`, `id_rsa*`, `id_ed25519*`
- - Credential folders: `secrets/`, `credentials/`, `private/`, `~/.ssh/`, `~/.aws/`
- - Agent session tokens: `~/.claude/auth*`, `~/.cursor/**`
- If user inputs contain credentials, you must invoke `scripts/secret-scan.js` or substitute with `[REDACTED:<TYPE>]`.
- 3. **Filesystem Scope Boundary**: Output writes are strictly confined to `./docs/` and installer backups in `~/.architect-skill-backup/`. You must never write outside these directories.
- 4. **No Guesswork / No AI Hallucinations**: Every architecture decision must cite standard principles (Clean Architecture, Modular Monolith, EDA, Hexagonal, FSD) and authoritative requirement standards (Atlassian PRD, BABOK BRD, ISO/IEC/IEEE 29148:2018 SRD).
- 5. **Context & Token Economy (High-Level vs Deep Analysis)**:
- - **Trivial / Non-Technical Queries**: Deliver an immediate, high-level summary and concise conclusion. Minimize context and token overhead; omit boilerplate.
- - **Heavy Technical / System Architecture Tasks**: Conduct rigorous first-principles analysis until the optimal solution is reached. Always present the analysis and obtain explicit developer approval before mutating files or executing system changes.
- ---
- ## Operational Workflow Overview
- The skill operates across two distinct tracks:
- - **Tier 1 (Rapid / Lightweight Architecture)**: Designed for MVPs, microservices, prototypes, or developer tools. Executes in 3 to 5 steps, generating core `PRD.md`, `layer-rules.md`, `directory-tree.md`, 1x ADR, C4 Container & ERD diagrams, UI setup guide, and theme tokens.
- - **Tier 2 (Enterprise / Comprehensive Architecture)**: Designed for mission-critical, regulated, or high-scale distributed systems. Executes in 6 deep validation stages, generating formal `BRD.md`, `PRD.md`, `SRD.md`, comprehensive ADRs, Redis caching lifecycle, distributed locking with fencing tokens, RDBMS scaling, 5 Mermaid UML diagrams, and full UI component registries with WCAG accessibility.
- ```
- [User Request / Intent Detected]
- |
- v
- [Phase 1: Triage & State Inspection]
- |
- +-------------+-------------+
- | |
- [Fast-Path Query] [Architecture Project]
- | |
- (Answer direct v
- technical ref) [Phase 2: Tier Selection & Wizard]
- - Tier 1: Rapid Path (3-5 steps)
- - Tier 2: Enterprise Path (6 stages)
- |
- v
- [Phase 3: Artifact Generation]
- - Writes to ./docs/
- |
- v
- [Phase 4: Quality Gate & Handoff]
- - validate-mermaid.js
- - validate-output.js
- - secret-scan.js
- - Final manifest & audit log
- ```
- ---
- ## Phase 1: Context Detection, Triage, and State Inspection
- When invoked, immediately check if the user request is an isolated technical query or an architecture generation session:
- ### 1.1 Direct Technical Query & Non-Technical Triage (Fast-Path)
- - **Non-Technical / Trivial Questions**: Provide an immediate high-level summary and concise conclusion. Do NOT generate unnecessary text, filler, or boilerplate.
- - **Isolated Technical Queries** (e.g., "How does XFetch work in Redis?", "Explain Kleppmann fencing tokens", "Mermaid ERD syntax"):
- - Provide an immediate, precise answer drawing directly from the authoritative files in `references/` or `assets/uml/`.
- - Do NOT force the user into the full multi-stage wizard unless they intend to build or document a system.
- ### 1.2 State Inspection for Architecture Generation
- If the user requests system architecture, a new project design, or triggers `/architect`:
- 1. Check for an existing transient wizard session:
- - Check if `.architect-wizard-state.json` exists in `./docs/` or the root workspace.
- - If found and `status == "in_progress"`, offer to resume:
- *"I detected an incomplete architecture session for project '[project_name]' at Stage [current_stage]. Would you like to resume where you left off or reset and start fresh?"*
- 2. If no active session exists, initialize state tracking via:
- ```bash
- node scripts/wizard-state.js init --tier [tier1|tier2] --project [project-name]
- ```
- ### 1.3 Mid-Flight Pivot & Revision Protocol (Changing Architecture or UI Mid-Way)
- If at any point during or after the wizard the user is not satisfied with the architectural style, wants to switch diagrams, changes database strategy, or requests a different UI framework / visual aesthetic:
- 1. **Zero Frustration / Non-Destructive Pivot**:
- - Never force the user to start completely from scratch or re-answer unaffected questions (e.g., changing UI aesthetic from Neubrutalism to Bento Grid does NOT require re-entering PRD requirements or database design).
- 2. **Selective State Rollback**:
- - Invoke `scripts/wizard-state.js pivot` to update state target and rewind `current_stage`:
- ```bash
- # To change UI framework or visual aesthetic mid-way:
- node scripts/wizard-state.js pivot --aesthetic bento-grid --ui-library tailwind-shadcn
- # To change core architectural style or diagram model mid-way:
- node scripts/wizard-state.js pivot --architecture modular-monolith --stage 2
- ```
- 3. **Artifact Regeneration & Differential Cleanup**:
- - Regenerate only the target subsystem files (`./docs/ui/` or `./docs/architecture/` or `./docs/diagrams/`).
- - Clean up deprecated twin files or old aesthetic tokens cleanly.
- - Re-run validation gates (`node scripts/validate-mermaid.js` and `node scripts/validate-ui-slop.js`).
- ---
- ## Phase 2: Interactive Wizard Protocol
- Guide the user through the structured questionnaire. Present clear options, concise trade-off explanations, and recommended defaults.
- ### Stage 0: System Identity & Tier Selection
- #### Prompt P0.1: Architecture Tier
- *"Choose the architecture tier for this project:*
- 1. **Tier 1 (Rapid / Lightweight)**: Optimal for MVPs, startups, prototypes, or microservices. Concise PRD, core architecture rules, C4/ERD diagrams, and essential UI tokens (< 5 minutes).
- 2. **Tier 2 (Enterprise / Comprehensive)**: Optimal for mission-critical, multi-team, or regulated systems. BABOK BRD, Atlassian PRD, ISO/IEC/IEEE 29148:2018 SRD, distributed locking with fencing tokens, Redis caching lifecycle, RDBMS scaling, 5 Mermaid UML diagrams, and component registries."*
- #### Prompt P0.2: Project Scope & Core Identity
- Ask for:
- - **Project Name**: (kebab-case, e.g., `settlement-engine`)
- - **Business Domain**: (e.g., Fintech, Healthcare, E-Commerce, Logistics, SaaS, Developer Tools)
- - **Target Platform**: (Web SPA, Mobile iOS/Android, Headless API, Distributed Worker, CLI Utility)
- - **Primary Tech Stack**: (Language, Runtime, Primary Database)
- ---
- ### Stage 1: Requirements Elicitation
- #### Fast-Track Auto-Ingestion (Existing PRD / BRD / SRD):
- - **Precondition**: If the user provides, attaches, points to, or already has existing requirements files (`PRD`, `BRD`, `SRD`, or any subset/single document among them):
- 1. **Immediate Ingestion**: Read the provided requirement document(s) directly using your file-reading tool.
- 2. **Zero-Prompt Bypass**: Do NOT ask the user whether they want to create/generate PRD, BRD, or SRD, and do NOT force them to choose options.
- 3. **Direct Progression**: Ingest the provided context and immediately advance straight to **Stage 2: Architectural Paradigm & Structural Boundaries**. If only one document is provided (e.g., only PRD), derive the architectural constraints directly from it without forcing the creation of missing documents.
- #### Standard Elicitation (When No Prior Requirements Are Provided):
- #### If Tier 1:
- Collect:
- 1. **Problem Statement**: What core pain point does this system eliminate?
- 2. **Target Personas**: Who are the primary end users?
- 3. **Core Capabilities**: List 3 to 5 must-have functional capabilities.
- *Emit*: `./docs/requirements/PRD.md` (Lightweight version with Gherkin acceptance criteria).
- #### If Tier 2:
- Collect:
- 1. **Business Justification (BRD - BABOK Guide v3)**:
- - Strategic business objectives, financial ROI targets, and operational metrics.
- - Stakeholder RACI matrix (Responsible, Accountable, Consulted, Informed).
- - Regulatory and compliance constraints (GDPR, HIPAA, PCI-DSS, SOC 2).
- *Emit*: `./docs/requirements/BRD.md`.
- 2. **Product Definition (PRD - Atlassian Methodology)**:
- - Detailed user personas and end-to-end user journeys.
- - Functional epics with Gherkin scenarios (`Given... When... Then...`).
- - MoSCoW feature prioritization and non-functional latency/availability targets.
- *Emit*: `./docs/requirements/PRD.md`.
- 3. **Formal System Requirements (SRD - ISO/IEC/IEEE 29148:2018)**:
- - Rigorous requirement clauses following EARS syntax:
- - Ubiquitous: *The [system] shall [action].*
- - Event-driven: *When [event], the [system] shall [action].*
- - State-driven: *While [in state], the [system] shall [action].*
- - Unwanted event: *If [error], then the [system] shall [action].*
- - Optional: *Where [feature enabled], the [system] shall [action].*
- - Independent Verification & Validation (IV&V) matrix with methods: `[I]` Inspection, `[A]` Analysis, `[D]` Demonstration, `[T]` Test.
- *Emit*: `./docs/requirements/SRD.md`.
- ---
- ### Stage 2: Architectural Paradigm & Structural Boundaries
- #### Prompt P2.1: Architectural Style
- Present the architectural paradigms:
- 1. **Modular Monolith**:
- - Single deployment unit, strictly encapsulated bounded contexts, internal public API facades, isolated database schemas.
- - Recommended for: Fast development velocity, evolving domain boundaries, teams seeking to defer microservice operational complexity.
- 2. **Clean Architecture / Hexagonal (Ports & Adapters)**:
- - Uncle Bob's 4 concentric layers (Entities, Use Cases, Interface Adapters, Frameworks).
- - Strict inward Dependency Rule. Frameworks and databases are swappable plugins.
- - Recommended for: Core enterprise business logic requiring unit testability without running databases.
- 3. **Event-Driven Architecture (EDA)**:
- - Asynchronous messaging via Kafka, RabbitMQ, or AWS SQS.
- - Domain Events, Event-Carried State Transfer (ECST), Transactional Outbox pattern, and Saga choreography/orchestration.
- - Recommended for: High-throughput ingestion, distributed scalability, decoupled workflows.
- 4. **Feature-Sliced Design (FSD v2.1)**:
- - Frontend and full-stack architecture structured into strict downward hierarchical layers: `app`, `pages`, `widgets`, `features`, `entities`, `shared`.
- - Recommended for: Scalable frontend applications with multiple product squads.
- #### Stage 2 Artifacts:
- - `./docs/architecture/decisions/ADR-001-architecture-style.md` (Michael Nygard format: Status, Context, Decision, Consequences).
- - `./docs/architecture/[selected-architecture]/layer-rules.md` (Strict layer boundary and import rules).
- - `./docs/architecture/[selected-architecture]/directory-tree.md` (Concrete folder hierarchy with module responsibilities).
- - `./docs/architecture/[selected-architecture]/dependency-injection.md` (Tier 2 only: Service lifetimes, DI container patterns, factory abstractions).
- ---
- ### Stage 3: Distributed System Design & Data Modeling
- #### Prompt P3.1: Data Storage & Persistence
- - Primary Database Engine: PostgreSQL, MySQL, MongoDB, DynamoDB, SQLite.
- - Schema isolation and partitioning strategy.
- #### Prompt P3.2: Distributed Resilience Patterns (Tier 2 mandatory):
- 1. **Redis Caching Strategy**:
- - Cache pattern: Cache-Aside, Write-Through, or Write-Behind.
- - Invalidation rules, TTL jitter, and eviction policy (`allkeys-lru`).
- - Stampede prevention: Probabilistic early expiration (XFetch algorithm) or singleflight mutex locks.
- *Emit*: `./docs/system-design/[design-name]/caching-lifecycle.md`.
- 2. **Concurrency & Distributed Locking**:
- - Mutual exclusion for critical resources (inventory, balances, state transitions).
- - Distributed locking with mandatory Kleppmann monotonically increasing fencing tokens to prevent split-brain writes during GC pauses or network delays.
- *Emit*: `./docs/system-design/[design-name]/concurrency-locking.md`.
- 3. **RDBMS Scaling & Message Queuing**:
- - Read/write replica splitting, connection pool sizing (PgBouncer/HikariCP), and table partitioning.
- - Message brokers with Dead Letter Queues (DLQ) and exponential backoff retry jitter.
- *Emit*: `./docs/system-design/[design-name]/rdbms-scaling.md` and `async-message-queue.md`.
- #### Stage 3 Diagrams (`./docs/diagrams/`):
- - `diagrams/architecture/c4-container.mmd` (Mermaid C4 Container model: Users, SPA, Gateway, Services, DB, Cache).
- - `diagrams/data/erd.mmd` (Mermaid ERD: Tables, Primary/Foreign Keys, 1:1, 1:N, N:M relationships, attributes).
- - Tier 2 additions:
- - `diagrams/sequence/sequence-[flow-name].mmd` (Auth flow, checkout flow, or async job flow).
- - `diagrams/state/state-[entity-name].mmd` (Order, invoice, or task state machine).
- - `diagrams/activity/activity-[workflow-name].mmd` (Business workflow with branch nodes).
- #### Preview-Safe Text Diagrams (Mandatory dual format):
- VSCode markdown preview does NOT render Mermaid. Every `.mmd` diagram MUST ship with a plain-text twin of the same base name, `diagrams/.../[name].md`, which renders everywhere:
- - Starts with a `# Title` heading, then a fenced ```` ```text ```` block with an ASCII diagram (use `+ - | > v` only), followed by a Markdown table (entities, transitions, or components).
- - Reference templates: `assets/diagrams/{sequence,erd,state,c4-container,activity}.md`. Mermaid stays as an optional extra for tools that render it.
- - Never put credentials in either format. Gate: `node scripts/validate-diagrams.js [md-dir] [mmd-dir]` must exit 0 (`validate-output.js` also warns when a twin is missing).
- ---
- ### Stage 4: Senior Frontend & Mobile Aesthetics (Web & Mobile UI)
- If the system has a user interface (Web, Mobile, Desktop):
- #### Prompt P4.1: UI Component Framework (Web & Mobile)
- Offer standard industry UI & Mobile component libraries:
- 1. **Web: Tailwind CSS + shadcn/ui** (Zero-runtime CSS, Radix UI accessibility, full local code ownership).
- 2. **Web: DaisyUI, Mantine, Chakra UI, Ant Design, or MUI** (Comprehensive enterprise web libraries).
- 3. **Mobile: React Native (Expo) + NativeWind / Tamagui** (Cross-platform iOS/Android sharing Tailwind tokens).
- 4. **Mobile: Flutter** (High-fidelity cross-platform custom canvas & animation engine).
- 5. **Mobile: Native iOS (SwiftUI) & Android (Jetpack Compose)** (Platform-native declarative UI with HIG/Material 3 standards).
- *Reference documentation: `assets/ui/ui-libraries.md` and `assets/ui/mobile-ui.md`.*
- #### Prompt P4.2: Senior Frontend Visual Aesthetics
- Offer 8 curated senior design aesthetic themes:
- 1. **Neubrutalism / Neo-Brutalism**: Hard 3px black borders, unblurred solid drop-shadows (`box-shadow: 4px 4px 0px #000`), bold pop palettes (Saweria, Gumroad).
- 2. **Bento Grid / Linear-style Minimalist**: Modular dark canvas (`#09090B`), subtle translucent borders, glowing ambient spotlight gradients (Linear, Raycast).
- 3. **Glassmorphism & Aurora Gradient**: Multi-layer frosted glass backdrop blurs (`backdrop-blur-xl`), vibrant celestial violet/cyan lighting.
- 4. **Swiss / International Typographic Style**: Stark architectural grid, bold grotesque typography, zero ornamentation, red/black accents.
- 5. **Modern SaaS Clean Enterprise**: Balanced slate neutrals, crisp 1px borders, subtle elevation, accessible status badges (Stripe, Vercel).
- 6. **Material Design 3**: Dynamic tonal palettes, surface tint elevations, pill-shaped action triggers.
- 7. **Neumorphism / Soft UI**: Opposing directional soft shadows on continuous neutral surfaces (with mandatory high-contrast text reinforcements).
- 8. **Retro / Cyberpunk / Terminal**: Monospaced typography, CRT scanlines, neon phosphor green/amber accents on pitch-black surfaces.
- #### Prompt P4.3: Anti-Slop & Strict Documentation-First Contract (Web & Mobile)
- Read `assets/ui/anti-slop.md` before generating or proposing any UI code. Strict Rules:
- - **Zero-Guesswork / Documentation-First Verification**:
- - Dilarang mengarang atau menebak sintaks komponen dari ingatan AI (Zero UI Hallucinations).
- - Untuk Mobile UI (React Native, Flutter, SwiftUI, Jetpack Compose, Tamagui, NativeWind), agent **WAJIB** memverifikasi sintaks, props, dan API resmi langsung dari dokumentasi terkini (via Context7 MCP untuk pustaka/SDK atau Web Search/Browser untuk dokumentasi resmi framework).
- - Untuk Web UI: Ambil kode sumber komponen/tema nyata via UI MCP yang terpasang (`neobrutalism`, `shadcn-ui`, `magicui`, `ui-registry`, `tailwindcss`) dan catat di `reference-log.md`.
- - **User Tech Flexibility**:
- - Pilihan teknologi mobile sepenuhnya menyesuaikan preferensi dan stack yang diinginkan user (React Native/Expo, Flutter, SwiftUI, Jetpack Compose, Kotlin Multiplatform, dll.).
- - **One Token Source (`theme-tokens.md`)**:
- - Dilarang menggunakan nilai warna/border sembarangan (*hardcoded arbitrary values*); seluruh elemen UI wajib merujuk pada token tema yang telah disepakati.
- - **Verification Gate**:
- - `node scripts/validate-ui-slop.js <ui-src-dir>` must exit 0.
- #### Stage 4 Artifacts:
- - `./docs/ui/[framework-name]/reference-log.md` (MCP calls and components used, per anti-slop contract).
- - `./docs/ui/[framework-name]/setup-guide.md` (Installation, tailwind config, build setup).
- - `./docs/ui/[framework-name]/theme-tokens.md` (CSS custom properties, exact colors, borders, shadows, typography).
- - `./docs/ui/[framework-name]/component-registry.md` (Core components: Button, Input, Modal, Table, Card with ARIA attributes and WCAG 2.1 AA contrast).
- ---
- ### Stage 5: Artifact Synthesis, Validation Gates, and Final Handoff
- #### 5.1 Compile Core Index & Manifest Files:
- 1. Generate `./docs/README.md`: Root navigation index linking all generated requirements, architecture, system design, diagrams, and UI documents.
- 2. Generate `./docs/.architect-manifest.json`: Machine-readable manifest recording generator version, project metadata, option selections, generated file paths, and SHA-256 hashes.
- 3. Generate `./docs/.architect-audit.log`: Local append-only event log recording timestamps and file generation actions (scrubbed of any sensitive data).
- 4. Generate `./docs/.gitignore`: Excluding `.env`, `*.pem`, `*.key`, `secrets/`, and `.architect-wizard-state.json`.
- #### 5.2 Execute Quality Gate Verification:
- You must run the local verification suite to ensure 100% integrity:
- ```bash
- # 1. Validate syntax of all generated Mermaid diagrams
- node scripts/validate-mermaid.js
- # 1b. Validate the preview-safe text twin of every diagram (renders in VSCode preview)
- node scripts/validate-diagrams.js
- # 2. Validate structural and markdown link integrity
- node scripts/validate-structure.js
- # 3. Validate cross-document semantic consistency (PRD <-> SRD <-> ERD <-> ADR)
- node scripts/validate-output.js
- # 4. Scan generated ./docs/ for stray secrets or sensitive keys
- node scripts/secret-scan.js --check
- ```
- If any validation step fails, immediately inspect the failure, repair the defective artifact, and re-run verification until all quality gates pass cleanly.
- #### 5.3 Finalize State and Deliver Handoff:
- Execute:
- ```bash
- node scripts/wizard-state.js finalize
- ```
- Present the user with:
- - Executive Summary of the generated architecture.
- - Directory map of created files in `./docs/`.
- - Concrete next steps for the engineering team (kickoff plan, code scaffolding, and CI integration).
README.md · README · 445 lines
- # architect-skill
- > Enterprise-grade, local-first, zero-egress software architecture generator and specification framework for AI agents.
- [](SECURITY.md)
- [](LICENSE)
- [](docs/workflow.md)
- [](CONTRIBUTING.md)
- [](https://nodejs.org)
- [](adapters/README.md)
- ---
- <p align="center">
- <a href="docs/workflow.md"><b>📖 Dokumentasi Alur (Workflow)</b></a> •
- <a href="docs/quickstart.md"><b>🚀 Quickstart Guide</b></a> •
- <a href="docs/wizard-guide.md"><b>🧙 Wizard Guide</b></a> •
- <a href="adapters/README.md"><b>🤖 Multi-Agent Adapters</b></a> •
- <a href="CONTRIBUTING.md"><b>🤝 Contributing</b></a> •
- <a href="SECURITY.md"><b>🛡️ Security Policy</b></a> •
- <a href="LICENSE"><b>📄 MIT License</b></a>
- </p>
- ---
- ## 1. Overview & Core Philosophies
- `architect-skill` is an authoritative software architecture toolkit designed to operate across seven major AI coding environments: Claude Code & Desktop, Cursor IDE, Continue.dev, Codeium Windsurf, Cline Autonomous Agent, OpenAI Codex CLI, and Google Antigravity.
- Unlike typical AI prompts that produce superficial code suggestions or make ungrounded assumptions, `architect-skill` operates as a deterministic architectural inference engine. It synthesizes enterprise-grade specifications, data schemas, distributed system safeguards, and visual UI tokens into an organized `./docs/` hierarchy within your project.
- ### Foundational Principles:
- 1. **Primacy of Logic over Heuristics**: All architectural paradigms, system design trade-offs, and requirement clauses are rooted in formal standards (ISO/IEC/IEEE 29148:2018, BABOK Guide v3, Atlassian Agile PRD, Robert C. Martin, Martin Fowler, Martin Kleppmann).
- 2. **Existing Requirements Fast-Track**: If a user brings an existing PRD, BRD, or SRD (or any subset), the system auto-ingests the documents directly and bypasses Stage 1 to proceed to Stage 2 architecture modeling without redundant elicitation questions.
- 3. **Dual-Track Context & Token Economy**: Trivial, high-level, and non-technical questions are answered concisely with immediate conclusions to save tokens and context window. Heavy technical challenges undergo deep first-principles analysis until resolved, followed by an explicit developer approval gate before files or systems are modified.
- 4. **Absolute Zero-Egress & Local-First**: 100% offline execution. No outbound network requests, no remote telemetry, no external API dependence. All memories, vibe-coding artifacts, and state are strictly persisted locally, eliminating cloud training risks.
- 5. **Absolute Credential Protection**: The skill is structurally prohibited from reading `.env` files, system credentials, SSH/AWS keys, certificates, private keys, or agent authentication tokens. Built-in secret scanning and automatic masking neutralize stray credentials before write operations.
- 6. **Constructive Minimalism**: Pure Node.js implementation without external npm dependencies (`zero-dependency`), eliminating software supply chain risks.
- ### End-to-End Architectural Workflow Diagram
- ```mermaid
- flowchart TD
- subgraph Phase1 ["1. Input & Elicitation Gate"]
- START([User / Developer Prompt]) --> CHK_REQ{Existing Requirements?<br/>PRD / BRD / SRD}
- CHK_REQ -- Yes: Fast-Track --> FAST_TRACK[Auto-Ingest Requirements Directly<br/>Bypass Stage 1 directly to Architecture]
- CHK_REQ -- No: Interactive --> STAGE1[Stage 1: Formal Requirements Elicitation<br/>- PRD: Atlassian Agile<br/>- BRD: BABOK Guide v3<br/>- SRD: ISO/IEC/IEEE 29148 EARS]
- end
- subgraph Governance ["Dual-Track Token Economy & Approval Gate"]
- GOV_CHK{Query Type}
- GOV_CHK -- Trivial / Non-Technical --> QUICK_RESP[Concise High-Level Summary]
- GOV_CHK -- Heavy Technical Challenge --> DEEP_ANALYSIS[Deep First-Principles Analysis]
- DEEP_ANALYSIS --> DEV_APPROVE{Developer Approval?}
- DEV_APPROVE -- Revise --> DEEP_ANALYSIS
- DEV_APPROVE -- Approved --> STAGE2
- end
- FAST_TRACK --> STAGE2[Stage 2: Architectural Paradigm & Boundaries<br/>- Clean Architecture & Hexagonal<br/>- Modular Monolith & FSD v2.1<br/>- Event-Driven Architecture<br/>- Michael Nygard ADRs]
- STAGE1 --> STAGE2
- subgraph Phase2 ["2. Engineering Design & Modeling"]
- STAGE2 --> STAGE3[Stage 3: Distributed System Design<br/>- Kleppmann Monotonic Fencing Tokens<br/>- Redis XFetch Anti-Stampede Caching<br/>- RDBMS Read/Write Split & Pool Sizing<br/>- Kafka/RabbitMQ Async Queues]
- STAGE3 --> STAGE4[Stage 4: Senior Frontend & UI Frameworks<br/>- 6 Frameworks: shadcn, DaisyUI, Mantine, etc.<br/>- 8 Aesthetics: Bento Grid, Neubrutalism, etc.]
- STAGE4 --> STAGE5[Stage 5: Visual UML & Mermaid Modeling<br/>- C4 Container Architecture<br/>- Sequence, ERD, State, & Activity Flow]
- end
- subgraph Phase3 ["3. Automated Verification & Isolation Gate"]
- STAGE5 --> STAGE6[Stage 6: Output Verification & Sandboxing<br/>- Zero-Egress Network Neutralization<br/>- Secret Scanner: Masking .env & Token<br/>- SHA-256 Tool Pinning Verification]
- STAGE6 --> CHK_VALID{All Checks Pass?}
- CHK_VALID -- Fail --> REMEDIATE[Auto-Scrub & Remediate Artifacts]
- REMEDIATE --> STAGE6
- CHK_VALID -- Pass --> ARTIFACTS([Compiled ./docs/ Output Tree<br/>100% Clean, Traceable & Production-Ready])
- end
- ```
- ---
- ## 2. Two-Tier Execution Framework
- `architect-skill` provides two calibrated operational tiers:
- | Attribute | Tier 1: Rapid / Lightweight Architecture | Tier 2: Enterprise / Comprehensive Architecture |
- |---|---|---|
- | **Target Scope** | MVPs, early prototypes, standalone microservices, internal CLI utilities | Mission-critical systems, distributed clusters, regulated enterprise platforms |
- | **Interaction Time** | 3 to 5 rapid steps (< 5 minutes) | 6 deep validation stages (15 to 30 minutes) |
- | **Requirements Scope** | Consolidated `PRD.md` with problem framing and acceptance criteria | Complete `BRD.md` (BABOK RACI/ROI), `PRD.md` (Atlassian), and `SRD.md` (ISO 29148 EARS clauses) |
- | **Architecture Scope** | `layer-rules.md`, `directory-tree.md`, 1x core ADR | Layer rules, directory tree, `dependency-injection.md`, 3+ comprehensive ADRs |
- | **System Design** | Primary database schema selection and baseline caching | Redis caching lifecycle, distributed locking with fencing tokens, RDBMS scaling, message queues |
- | **Diagrams (Mermaid + text twin)** | 2 diagrams: C4 Container and Relational ERD | 5 diagrams: C4 Container, Sequence, ERD, State Machine, Activity flow. Each ships as Mermaid `.mmd` plus an ASCII `.md` twin that renders in VSCode preview |
- | **Frontend & UI** | UI framework setup guide and theme tokens | Setup guide, theme tokens, full component registry with ARIA/WCAG accessibility |
- | **Verification Gate** | Basic syntax validation and manifest compilation | Cross-document semantic traceability verification (`validate-output.js`) |
- ---
- ## 3. Supported AI Agent Ecosystems
- `architect-skill` provides tailored adapters for 7 agent platforms:
- | Platform | Adapter File | Global Location | Project Location | Trigger Mechanism |
- |---|---|---|---|---|
- | **Claude Code & Desktop** | `adapters/claude/SKILL.md` | `~/.claude/skills/architect-skill/` | `<project>/.claude/skills/architect-skill/` | `/architect` |
- | **Cursor IDE** | `adapters/cursor/architect-skill.mdc` | `~/.cursor/rules/architect-skill.mdc` | `<project>/.cursor/rules/architect-skill.mdc` | Context / Globs |
- | **Continue.dev** | `adapters/continue/architect-skill.prompt` | `~/.continue/prompts/architect-skill.prompt` | `<project>/.continue/prompts/architect-skill.prompt` | `/architect` |
- | **Codeium Windsurf** | `adapters/windsurf/.windsurfrules` | `~/.codeium/windsurf/memories/global_rules.md` | `<project>/.windsurfrules` | Cascade Trigger |
- | **Cline Autonomous** | `adapters/cline/.clinerules` | Extension configuration | `<project>/.clinerules` | Autonomous Rules |
- | **OpenAI Codex CLI** | `adapters/codex/SKILL.md` | `~/.codex/skills/architect-skill/` | `<project>/.codex/skills/architect-skill/` | CLI Command |
- | **Google Antigravity** | `adapters/antigravity/SKILL.md` | `~/.agy/skills/architect-skill/` | `<project>/.agy/skills/architect-skill/` | Skill Loader |
- Detailed installation mapping is documented in [`adapters/README.md`](adapters/README.md).
- ---
- ## 4. Installation & Configuration
- The toolkit includes a zero-dependency installer and configuration engine compatible with Linux, macOS, and Windows across all 7 supported AI agents.
- ### 4.1 Zero-Clone Installation via npm & npx
- The package is officially published on the NPM Registry (`npm i skill-architecture`). Users do not need to clone this repository. You can install or run `skill-architecture` across all AI agents with a single command:
- #### Standard CLI Execution via `npx` (No Install Required):
- ```bash
- # Install globally to all supported AI agents in one command
- npx skill-architecture --global
- # Target a specific AI agent directly (e.g., Antigravity, Cursor, or Claude)
- npx skill-architecture --global --agent antigravity
- npx skill-architecture --global --agent cursor
- npx skill-architecture --global --agent claude
- # Local workspace installation
- npx skill-architecture --local --target all
- ```
- #### Install via `npm` (Global or Project Dependency):
- ```bash
- # Install globally on your machine
- npm install -g skill-architecture
- skill-architecture --global
- # Or install as a dev dependency in your project
- npm install -D skill-architecture
- # or simply:
- npm i skill-architecture
- ```
- #### Industry Standard via `npx skills` (skills.sh ecosystem by Vercel Labs):
- Because this repository conforms strictly to the open `SKILL.md` specification, users of the universal agent skill manager can also install it natively:
- ```bash
- # Add to all supported agents globally
- npx skills add skill-architecture -g
- # or from GitHub:
- npx skills add FIKY0911/skill-architecture -g
- # Add to a specific agent (e.g., Cursor or Claude Code)
- npx skills add skill-architecture --agent cursor
- npx skills add skill-architecture --agent claude-code
- ```
- ---
- ### 4.2 Direct ZIP Package Upload (Zero-Terminal / Web AI UI)
- For users who want to use this skill directly inside web-based AI interfaces without using a terminal or command line:
- 1. **Download the standalone ZIP package**:
- - Download [`architect-skill.zip`](file:///home/fiky/Documents/Latihan Agent/SKILL/architect-skill/architect-skill.zip) directly from the repository / GitHub Releases or build it locally using `npm run bundle:zip`.
- 2. **Upload to Your AI Agent Environment**:
- - **Claude Desktop / Claude.ai Projects**: Attach or upload `architect-skill.zip` to your project knowledge base or custom instructions.
- - **Google Antigravity / Gemini Workspace**: Place the unzipped contents into `~/.gemini/skills/architect-skill` or upload via the skills manager.
- - **ChatGPT Custom GPT / Assistants**: Attach `architect-skill.zip` into the GPT's **Knowledge / Code Interpreter** storage.
- - **Cursor / VS Code IDE**: Unzip directly into your project's root folder.
- The ZIP bundle contains the complete offline architecture rulebook, deterministic state machine, zero-egress safeguards, and industry reference standards in a compact ~450 KB archive.
- ---
- ### 4.3 Local & Monorepo Installation (For Contributors)
- ```bash
- # Preview installation plan without touching disk (--dry-run)
- node scripts/install.js --global --dry-run
- # Install globally for all 7 AI agents
- node scripts/install.js --global
- # Install for a single specific AI agent
- node scripts/install.js --global --target antigravity
- node scripts/install.js --local --target cursor
- # Complete clean uninstallation
- node scripts/install.js --global --uninstall
- node scripts/install.js --local --uninstall
- ```
- ---
- ### 4.4 Recommended MCP Servers (Model Context Protocol)
- To unlock the full potential of `skill-architecture` (especially for pulling official library documentation without guessing and generating zero-slop UI components), install the following recommended MCP servers into your agent's configuration (`mcp_config.json`):
- #### 1. Core Documentation & Verification MCPs
- - **`context7` (`@upstash/context7-mcp`)**: **Essential.** Resolves up-to-date official documentation and API references for libraries, frameworks, and SDKs (React, Next.js, Flutter, React Native, Tailwind, Redis, Prisma, etc.) directly in the prompt context.
- ```json
- "context7": {
- "command": "npx",
- "args": ["-y", "@upstash/context7-mcp"]
- }
- ```
- - **`filesystem` (`@modelcontextprotocol/server-filesystem`)**: Allows the agent to read workspace files and write architecture blueprints cleanly into `./docs/`.
- ```json
- "filesystem": {
- "command": "npx",
- "args": ["-y", "@modelcontextprotocol/server-filesystem", "/path/to/your/workspace"]
- }
- ```
- - **`fetch` (`mcp-server-fetch`)**: Fetches raw specifications, RFC standards, or web APIs offline or via trusted networks.
- ```json
- "fetch": {
- "command": "uvx",
- "args": ["mcp-server-fetch"]
- }
- ```
- #### 2. Design System & UI Engineering MCPs (Strict Documentation-First)
- These MCP servers enforce the **Anti-Slop UI Contract**, preventing the agent from hallucinating arbitrary CSS or mock UI components:
- - **`shadcn-ui` (`@jpisnice/shadcn-ui-mcp-server`)**: Standard accessible UI components and blocks.
- ```json
- "shadcn-ui": {
- "command": "npx",
- "args": ["-y", "@jpisnice/shadcn-ui-mcp-server"]
- }
- ```
- - **`neobrutalism` (`neobrutalism-mcp-server`)**: Neo-brutalist components, high-contrast borders, and pop palettes.
- ```json
- "neobrutalism": {
- "command": "npx",
- "args": ["-y", "neobrutalism-mcp-server"]
- }
- ```
- - **`magicui` (`@magicuidesign/mcp`)**: Interactive micro-interactions, landing page effects, and animations.
- ```json
- "magicui": {
- "command": "npx",
- "args": ["-y", "@magicuidesign/mcp"]
- }
- ```
- - **`ui-registry` (`ui-registry-mcp`)**: Catalog of 12+ modern design registries (Aceternity, Cult UI, SmoothUI, Kokonut).
- ```json
- "ui-registry": {
- "command": "npx",
- "args": ["-y", "ui-registry-mcp"]
- }
- ```
- - **`tailwindcss` (`tailwindcss-mcp`)**: Validates Tailwind CSS utility classes and theme configuration tokens.
- ```json
- "tailwindcss": {
- "command": "npx",
- "args": ["-y", "tailwindcss-mcp"]
- }
- ```
- #### 3. Browser Automation & Visual Testing MCPs
- - **`playwright` (`@playwright/mcp@latest`)**: Automated headless browser testing and visual screenshot validation.
- ```json
- "playwright": {
- "command": "npx",
- "args": ["-y", "@playwright/mcp@latest"]
- }
- ```
- - **`puppeteer` (`@modelcontextprotocol/server-puppeteer`)**: Headless browser automation and pixel-perfect PDF document rendering.
- ```json
- "puppeteer": {
- "command": "npx",
- "args": ["-y", "@modelcontextprotocol/server-puppeteer"]
- }
- ```
- ---
- ### 4.5 Agent Configuration Mapping
- When installed, each agent consumes the skill via its native configuration architecture:
- | Target Agent | Installation Location | Integration Mechanism |
- | :--- | :--- | :--- |
- | **Google Antigravity (AGY)** | `~/.gemini/skills/architect-skill` & `~/.agy/skills/` | Loaded as native agent skill & callable via prompt/commands |
- | **Claude Desktop / Code** | `~/.claude/skills/architect-skill/SKILL.md` | Declarative markdown skill injected into agent context |
- | **Cursor IDE** | `~/.cursor/rules/architect-skill.mdc` | Semantic context rule with glob pattern indexing |
- | **Continue.dev** | `~/.continue/prompts/architect-skill.prompt` | Custom architecture design prompt template |
- | **Windsurf** | `~/.codeium/windsurf/.windsurfrules` or workspace `.windsurfrules` | System-level architectural directives |
- | **Cline** | Workspace `.clinerules` | Autonomous agent behavioral constraints |
- | **OpenAI Codex** | `~/.codex/skills/architect-skill` | Skill package for multi-turn architectural codegen |
- ### 4.4 Automated Backup & Rollback Protection
- Prior to modifying or overwriting any agent configuration, the installer automatically creates a timestamped snapshot in:
- ```
- ~/.architect-skill-backup/<timestamp>/<agent_name>/
- ```
- To inspect or revert, simply restore files from the respective timestamp folder or execute `node scripts/install.js --uninstall`.
- ### 4.5 Execution Protocol (How to Run)
- 1. **Interactive Design Wizard**: Ask the agent:
- > *"Activate architect-skill to design a new system: [System Name/Domain]"*
- 2. **Deterministic Stages**: Follow the Tier 1/2 elicitation process (Requirements $\to$ Architecture Style $\to$ Frontend Aesthetics $\to$ System Design).
- 3. **Artifact Generation**: Review outputs automatically generated into `./docs/` (`docs/requirements/`, `docs/architecture/`, `docs/diagrams/`).
- 4. **Verification**: Run `npm test` and `npm run validate:mermaid` to ensure zero hallucinations and valid UML diagrams.
- For extensive guides and documentation, explore the [`docs/`](docs/) directory:
- - [Workflow Guide](docs/workflow.md)
- - [Quickstart & Configuration Details](docs/quickstart.md)
- - [Wizard Execution Guide](docs/wizard-guide.md)
- - [Multi-Agent Adapters Reference](adapters/README.md)
- - [Documentation Index & Specifications](docs/spec/README.id.md)
- - [Community & Security Policies](docs/community/CONTRIBUTING.md)
- ---
- ## 5. Architectural Knowledge Base & Standards
- ### 5.1 Formal Requirements Engineering
- - **PRD (Product Requirements Document)**: Follows Atlassian agile framework with problem framing, user personas, Gherkin acceptance criteria, and MoSCoW prioritization.
- - **BRD (Business Requirements Document)**: Follows BABOK Guide v3 with strategic alignment, stakeholder RACI matrices, TCO/ROI financial calculations, and compliance constraints.
- - **SRD (System Requirements Document)**: Follows ISO/IEC/IEEE 29148:2018 (superseding IEEE 830) enforcing EARS syntax (Ubiquitous, Event-driven, State-driven, Unwanted, Optional) and IV&V verification matrices.
- ### 5.2 Architectural Paradigms
- - **Clean Architecture**: 4 concentric onion layers with the strict Dependency Inversion Rule and boundary DTOs.
- - **Modular Monolith**: Bounded contexts within a single deployment unit, public interface facades, and logical schema isolation.
- - **Event-Driven Architecture (EDA)**: Domain events, Event-Carried State Transfer (ECST), Transactional Outbox pattern, and Saga orchestration.
- - **Hexagonal Architecture**: Ports and Adapters decoupling application core from driving and driven infrastructure.
- - **Feature-Sliced Design (FSD v2.1)**: Strict hierarchical layer model (`app`, `pages`, `widgets`, `features`, `entities`, `shared`) for large-scale frontend and full-stack systems.
- ### 5.3 Distributed System Design Rigor
- - **Redis Caching**: Cache-Aside, Write-Through, Write-Behind, TTL jitter, and XFetch probabilistic early expiration algorithm to prevent cache stampedes.
- - **Distributed Locking**: Kleppmann monotonically increasing fencing tokens checked at the database layer to prevent split-brain writes under GC pauses.
- - **RDBMS Scaling**: Read/Write splitting with replica lag mitigation, PgBouncer/HikariCP connection pool sizing, and horizontal sharding.
- - **Async Message Queues**: Kafka, RabbitMQ, and AWS SQS patterns with Dead Letter Queues (DLQ) and exponential backoff jitter.
- - **Distributed Rate Limiting**: Sliding Window Counter and Token Bucket algorithms implemented via Redis Lua scripts.
- ### 5.4 Senior Frontend Aesthetics & UI Frameworks
- Integrates 6 official UI component libraries (Tailwind + shadcn/ui, DaisyUI, Mantine, Chakra UI, Ant Design, MUI) and 8 curated senior design aesthetic themes:
- 1. **Neubrutalism / Neo-Brutalism**: Hard 3px black borders, unblurred solid drop-shadows (`box-shadow: 4px 4px 0px #000`), vibrant saturated accents.
- 2. **Bento Grid / Linear-style**: Modular dark canvas (`#09090B`), subtle translucent borders, glowing ambient spotlight gradients.
- 3. **Glassmorphism & Aurora Gradient**: Multi-layer frosted glass backdrop blurs (`backdrop-blur-xl`), vibrant celestial violet/cyan lighting.
- 4. **Swiss / International Typographic Style**: Stark architectural grid, bold grotesque typography, zero ornamentation, red/black accents.
- 5. **Modern SaaS Clean Enterprise**: Balanced slate neutrals, crisp 1px borders, subtle elevation, accessible status badges.
- 6. **Material Design 3**: Dynamic tonal palettes, surface tint elevations, pill-shaped action triggers.
- 7. **Neumorphism / Soft UI**: Opposing directional soft shadows on continuous neutral surfaces (with mandatory high-contrast text reinforcements).
- 8. **Retro / Cyberpunk / Terminal**: Monospaced typography, CRT scanlines, neon phosphor green/amber accents on pitch-black surfaces.
- ---
- ## 6. Generated Output Hierarchy (`./docs/`)
- When executed, `architect-skill` compiles your technical documentation into `./docs/`:
- ```
- ./docs/
- ├── README.md # Main navigation portal and documentation index
- ├── .architect-manifest.json # Machine-readable metadata, option choices, and SHA-256 hashes
- ├── .architect-audit.log # Local append-only event log (zero credential leakage)
- ├── .gitignore # Output-level exclusion of secrets, keys, and temporary state
- │
- ├── requirements/
- │ ├── PRD.md # Product Requirements Document (Atlassian)
- │ ├── BRD.md # Business Requirements Document (BABOK - Tier 2)
- │ ├── SRD.md # System Requirements Document (ISO/IEC/IEEE 29148 - Tier 2)
- │ └── [custom].md # Sanitized custom requirements or RFCs
- │
- ├── architecture/
- │ ├── decisions/
- │ │ ├── ADR-001-*.md # Architecture Decision Records
- │ │ └── ADR-002-*.md
- │ └── [selected-architecture]/
- │ ├── layer-rules.md # Boundary rules and import constraints
- │ ├── directory-tree.md # Concrete project codebase tree
- │ └── dependency-injection.md # Service lifetimes and DI registration (Tier 2)
- │
- ├── system-design/
- │ └── [selected-design]/
- │ ├── caching-lifecycle.md # Redis caching patterns and TTL policies
- │ ├── concurrency-locking.md # Distributed locking with fencing tokens
- │ └── rdbms-scaling.md # Read replicas, partitioning, and connection pools
- │
- ├── diagrams/
- │ ├── architecture/
- │ │ └── c4-container.mmd # C4 model container diagram
- │ ├── data/
- │ │ └── erd.mmd # Relational Entity Relationship Diagram
- │ ├── sequence/
- │ │ └── sequence-auth-flow.mmd # Protocol and transaction sequence diagrams
- │ └── state/
- │ └── state-lifecycle.mmd # Finite state machine model
- │
- └── ui/
- └── [selected-ui-framework]/
- ├── setup-guide.md # Build configuration and dependencies
- ├── theme-tokens.md # CSS custom properties and design tokens
- └── component-registry.md # UI component specifications and ARIA accessibility
- ```
- ---
- ## 7. Verification Utilities & Diagnostics
- The toolkit provides automated zero-dependency verification scripts:
- ```bash
- # Verify environment health, permissions, and script integrity
- node scripts/doctor.js
- # Validate Mermaid UML diagram syntax across assets and generated output
- node scripts/validate-mermaid.js
- # Validate preview-safe text diagrams (ASCII .md twins that render in VSCode preview without Mermaid)
- node scripts/validate-diagrams.js
- # Validate repository structure against the blueprint
- node scripts/validate-structure.js
- # Validate cross-document semantic consistency (PRD <-> SRD <-> ERD <-> ADR)
- node scripts/validate-output.js
- # Execute secret scanner to detect and mask credential patterns
- node scripts/secret-scan.js --check
- ```
- ---
- ## 8. Security & Governance
- - **Zero-Egress Enforcement**: Monitored by `scripts/egress-guard.js`. Any unauthorized outbound network socket throws an immediate fatal error.
- - **Credential Protection**: Monitored by `scripts/secret-scan.js` and `policies/redaction-rules.json`.
- - **Filesystem Isolation**: Monitored by `policies/filesystem-scope.json`.
- - **Integrity Pinning**: Monitored by `runtime-firewall/tool-firewall.js`.
- For security vulnerability disclosure, review [`SECURITY.md`](SECURITY.md).
- For local data sovereignty details, review [`PRIVACY.md`](PRIVACY.md).
- ---
- ## 9. License
- This project is licensed under the [MIT License](LICENSE).
Ratings and comments
Sign in to rate this skill.
Sign inFree to use. If it’s helped, pass it on.