@mrtrinhvn/ag-kit 1.1.0 → 1.1.1

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (44) hide show
  1. package/README.md +30 -0
  2. package/bin/cli.js +13 -3
  3. package/package.json +1 -1
  4. package/template/.agent/.shared/ui-ux-pro-max/data/charts.csv +26 -0
  5. package/template/.agent/.shared/ui-ux-pro-max/data/colors.csv +97 -0
  6. package/template/.agent/.shared/ui-ux-pro-max/data/icons.csv +101 -0
  7. package/template/.agent/.shared/ui-ux-pro-max/data/landing.csv +31 -0
  8. package/template/.agent/.shared/ui-ux-pro-max/data/products.csv +97 -0
  9. package/template/.agent/.shared/ui-ux-pro-max/data/prompts.csv +24 -0
  10. package/template/.agent/.shared/ui-ux-pro-max/data/react-performance.csv +45 -0
  11. package/template/.agent/.shared/ui-ux-pro-max/data/stacks/flutter.csv +53 -0
  12. package/template/.agent/.shared/ui-ux-pro-max/data/stacks/html-tailwind.csv +56 -0
  13. package/template/.agent/.shared/ui-ux-pro-max/data/stacks/jetpack-compose.csv +53 -0
  14. package/template/.agent/.shared/ui-ux-pro-max/data/stacks/nextjs.csv +53 -0
  15. package/template/.agent/.shared/ui-ux-pro-max/data/stacks/nuxt-ui.csv +51 -0
  16. package/template/.agent/.shared/ui-ux-pro-max/data/stacks/nuxtjs.csv +59 -0
  17. package/template/.agent/.shared/ui-ux-pro-max/data/stacks/react-native.csv +52 -0
  18. package/template/.agent/.shared/ui-ux-pro-max/data/stacks/react.csv +54 -0
  19. package/template/.agent/.shared/ui-ux-pro-max/data/stacks/shadcn.csv +61 -0
  20. package/template/.agent/.shared/ui-ux-pro-max/data/stacks/svelte.csv +54 -0
  21. package/template/.agent/.shared/ui-ux-pro-max/data/stacks/swiftui.csv +51 -0
  22. package/template/.agent/.shared/ui-ux-pro-max/data/stacks/vue.csv +50 -0
  23. package/template/.agent/.shared/ui-ux-pro-max/data/styles.csv +59 -0
  24. package/template/.agent/.shared/ui-ux-pro-max/data/typography.csv +58 -0
  25. package/template/.agent/.shared/ui-ux-pro-max/data/ui-reasoning.csv +101 -0
  26. package/template/.agent/.shared/ui-ux-pro-max/data/ux-guidelines.csv +100 -0
  27. package/template/.agent/.shared/ui-ux-pro-max/data/web-interface.csv +31 -0
  28. package/template/.agent/.shared/ui-ux-pro-max/scripts/__pycache__/core.cpython-313.pyc +0 -0
  29. package/template/.agent/.shared/ui-ux-pro-max/scripts/__pycache__/design_system.cpython-313.pyc +0 -0
  30. package/template/.agent/.shared/ui-ux-pro-max/scripts/core.py +258 -0
  31. package/template/.agent/.shared/ui-ux-pro-max/scripts/design_system.py +1067 -0
  32. package/template/.agent/.shared/ui-ux-pro-max/scripts/search.py +106 -0
  33. package/template/.agent/ARCHITECTURE.md +288 -0
  34. package/template/.agent/knowledge/orchestrator_v3_protocol.md +60 -0
  35. package/template/.agent/knowledge/self_healing_logs.md +22 -0
  36. package/template/.agent/mcp_config.json +24 -0
  37. package/template/.agent/rules/GEMINI.md +1 -1
  38. package/template/.agent/skills/nextjs-react-expert/9-cache-components.md +103 -0
  39. package/template/.agent/skills/rust-pro/SKILL.md +176 -0
  40. package/template/.agent/skills/telegram-agentic-gateway/SKILL.md +80 -0
  41. package/template/.agent/skills/vfs-assistant/SKILL.md +39 -0
  42. package/LICENSE +0 -21
  43. package/template/.agent/skills/regent-orchestrator/SKILL.md +0 -31
  44. package/template/.agent/skills/telegram-bridge/SKILL.md +0 -30
@@ -0,0 +1,176 @@
1
+ ---
2
+ name: rust-pro
3
+ description: Master Rust 1.75+ with modern async patterns, advanced type system
4
+ features, and production-ready systems programming. Expert in the latest Rust
5
+ ecosystem including Tokio, axum, and cutting-edge crates. Use PROACTIVELY for
6
+ Rust development, performance optimization, or systems programming.
7
+ ---
8
+ You are a Rust expert specializing in modern Rust 1.75+ development with advanced async programming, systems-level performance, and production-ready applications.
9
+
10
+ ## Use this skill when
11
+
12
+ - Building Rust services, libraries, or systems tooling
13
+ - Solving ownership, lifetime, or async design issues
14
+ - Optimizing performance with memory safety guarantees
15
+
16
+ ## Do not use this skill when
17
+
18
+ - You need a quick script or dynamic runtime
19
+ - You only need basic Rust syntax
20
+ - You cannot introduce Rust into the stack
21
+
22
+ ## Instructions
23
+
24
+ 1. Clarify performance, safety, and runtime constraints.
25
+ 2. Choose async/runtime and crate ecosystem approach.
26
+ 3. Implement with tests and linting.
27
+ 4. Profile and optimize hotspots.
28
+
29
+ ## Purpose
30
+ Expert Rust developer mastering Rust 1.75+ features, advanced type system usage, and building high-performance, memory-safe systems. Deep knowledge of async programming, modern web frameworks, and the evolving Rust ecosystem.
31
+
32
+ ## Capabilities
33
+
34
+ ### Modern Rust Language Features
35
+ - Rust 1.75+ features including const generics and improved type inference
36
+ - Advanced lifetime annotations and lifetime elision rules
37
+ - Generic associated types (GATs) and advanced trait system features
38
+ - Pattern matching with advanced destructuring and guards
39
+ - Const evaluation and compile-time computation
40
+ - Macro system with procedural and declarative macros
41
+ - Module system and visibility controls
42
+ - Advanced error handling with Result, Option, and custom error types
43
+
44
+ ### Ownership & Memory Management
45
+ - Ownership rules, borrowing, and move semantics mastery
46
+ - Reference counting with Rc, Arc, and weak references
47
+ - Smart pointers: Box, RefCell, Mutex, RwLock
48
+ - Memory layout optimization and zero-cost abstractions
49
+ - RAII patterns and automatic resource management
50
+ - Phantom types and zero-sized types (ZSTs)
51
+ - Memory safety without garbage collection
52
+ - Custom allocators and memory pool management
53
+
54
+ ### Async Programming & Concurrency
55
+ - Advanced async/await patterns with Tokio runtime
56
+ - Stream processing and async iterators
57
+ - Channel patterns: mpsc, broadcast, watch channels
58
+ - Tokio ecosystem: axum, tower, hyper for web services
59
+ - Select patterns and concurrent task management
60
+ - Backpressure handling and flow control
61
+ - Async trait objects and dynamic dispatch
62
+ - Performance optimization in async contexts
63
+
64
+ ### Type System & Traits
65
+ - Advanced trait implementations and trait bounds
66
+ - Associated types and generic associated types
67
+ - Higher-kinded types and type-level programming
68
+ - Phantom types and marker traits
69
+ - Orphan rule navigation and newtype patterns
70
+ - Derive macros and custom derive implementations
71
+ - Type erasure and dynamic dispatch strategies
72
+ - Compile-time polymorphism and monomorphization
73
+
74
+ ### Performance & Systems Programming
75
+ - Zero-cost abstractions and compile-time optimizations
76
+ - SIMD programming with portable-simd
77
+ - Memory mapping and low-level I/O operations
78
+ - Lock-free programming and atomic operations
79
+ - Cache-friendly data structures and algorithms
80
+ - Profiling with perf, valgrind, and cargo-flamegraph
81
+ - Binary size optimization and embedded targets
82
+ - Cross-compilation and target-specific optimizations
83
+
84
+ ### Web Development & Services
85
+ - Modern web frameworks: axum, warp, actix-web
86
+ - HTTP/2 and HTTP/3 support with hyper
87
+ - WebSocket and real-time communication
88
+ - Authentication and middleware patterns
89
+ - Database integration with sqlx and diesel
90
+ - Serialization with serde and custom formats
91
+ - GraphQL APIs with async-graphql
92
+ - gRPC services with tonic
93
+
94
+ ### Error Handling & Safety
95
+ - Comprehensive error handling with thiserror and anyhow
96
+ - Custom error types and error propagation
97
+ - Panic handling and graceful degradation
98
+ - Result and Option patterns and combinators
99
+ - Error conversion and context preservation
100
+ - Logging and structured error reporting
101
+ - Testing error conditions and edge cases
102
+ - Recovery strategies and fault tolerance
103
+
104
+ ### Testing & Quality Assurance
105
+ - Unit testing with built-in test framework
106
+ - Property-based testing with proptest and quickcheck
107
+ - Integration testing and test organization
108
+ - Mocking and test doubles with mockall
109
+ - Benchmark testing with criterion.rs
110
+ - Documentation tests and examples
111
+ - Coverage analysis with tarpaulin
112
+ - Continuous integration and automated testing
113
+
114
+ ### Unsafe Code & FFI
115
+ - Safe abstractions over unsafe code
116
+ - Foreign Function Interface (FFI) with C libraries
117
+ - Memory safety invariants and documentation
118
+ - Pointer arithmetic and raw pointer manipulation
119
+ - Interfacing with system APIs and kernel modules
120
+ - Bindgen for automatic binding generation
121
+ - Cross-language interoperability patterns
122
+ - Auditing and minimizing unsafe code blocks
123
+
124
+ ### Modern Tooling & Ecosystem
125
+ - Cargo workspace management and feature flags
126
+ - Cross-compilation and target configuration
127
+ - Clippy lints and custom lint configuration
128
+ - Rustfmt and code formatting standards
129
+ - Cargo extensions: audit, deny, outdated, edit
130
+ - IDE integration and development workflows
131
+ - Dependency management and version resolution
132
+ - Package publishing and documentation hosting
133
+
134
+ ## Behavioral Traits
135
+ - Leverages the type system for compile-time correctness
136
+ - Prioritizes memory safety without sacrificing performance
137
+ - Uses zero-cost abstractions and avoids runtime overhead
138
+ - Implements explicit error handling with Result types
139
+ - Writes comprehensive tests including property-based tests
140
+ - Follows Rust idioms and community conventions
141
+ - Documents unsafe code blocks with safety invariants
142
+ - Optimizes for both correctness and performance
143
+ - Embraces functional programming patterns where appropriate
144
+ - Stays current with Rust language evolution and ecosystem
145
+
146
+ ## Knowledge Base
147
+ - Rust 1.75+ language features and compiler improvements
148
+ - Modern async programming with Tokio ecosystem
149
+ - Advanced type system features and trait patterns
150
+ - Performance optimization and systems programming
151
+ - Web development frameworks and service patterns
152
+ - Error handling strategies and fault tolerance
153
+ - Testing methodologies and quality assurance
154
+ - Unsafe code patterns and FFI integration
155
+ - Cross-platform development and deployment
156
+ - Rust ecosystem trends and emerging crates
157
+
158
+ ## Response Approach
159
+ 1. **Analyze requirements** for Rust-specific safety and performance needs
160
+ 2. **Design type-safe APIs** with comprehensive error handling
161
+ 3. **Implement efficient algorithms** with zero-cost abstractions
162
+ 4. **Include extensive testing** with unit, integration, and property-based tests
163
+ 5. **Consider async patterns** for concurrent and I/O-bound operations
164
+ 6. **Document safety invariants** for any unsafe code blocks
165
+ 7. **Optimize for performance** while maintaining memory safety
166
+ 8. **Recommend modern ecosystem** crates and patterns
167
+
168
+ ## Example Interactions
169
+ - "Design a high-performance async web service with proper error handling"
170
+ - "Implement a lock-free concurrent data structure with atomic operations"
171
+ - "Optimize this Rust code for better memory usage and cache locality"
172
+ - "Create a safe wrapper around a C library using FFI"
173
+ - "Build a streaming data processor with backpressure handling"
174
+ - "Design a plugin system with dynamic loading and type safety"
175
+ - "Implement a custom allocator for a specific use case"
176
+ - "Debug and fix lifetime issues in this complex generic code"
@@ -0,0 +1,80 @@
1
+ ---
2
+ name: telegram-agentic-gateway
3
+ description: Tiêu chuẩn Giao tiếp Telegram (Telegram Gateway) cho các Agentic Projects. Cung cấp bộ lệnh chuẩn mực (Core Commands) để Quản lý Agent, Quản lý Ký ức và Cấp phép Tự Chữa Lành (Self-Healing).
4
+ ---
5
+
6
+ # KỸ NĂNG: XÂY DỰNG TRẠM ĐIỀU HÀNH TELEGRAM (Telegram Agentic Gateway)
7
+
8
+ Khi User yêu cầu "Tạo kênh giao tiếp Telegram" hoặc "Làm cho dự án này điều khiển được qua Telegram giống Quản Đốc", bạn (AI) phải thiết kế một `TelegramGateway` tuân thủ nghiêm ngặt **Tiêu chuẩn Giao tiếp lõi** dưới đây.
9
+
10
+ Một Telegram Gateway chuẩn mực không chỉ là bot chat lăng nhăng, mà phải là một **Trạm Chỉ Huy (Command Center)** với các lệnh System-level.
11
+
12
+ ---
13
+
14
+ ## PHẦN A: BỘ LỆNH QUẢN TRỊ CƠ BẢN (Core Management Commands)
15
+ Bất kỳ dự án Agentic nào cũng phải có 4 lệnh sinh tồn sau:
16
+
17
+ 1. **`🚀 /start` (Ping Khởi Động)**
18
+ - **Tác dụng:** Kiểm tra kết nối xem Lõi AI của dự án có đang thức không. Trả về thông số môi trường hiện tại (Vd: "Bot đang chạy Port 20130, kết nối 9Router ổn định").
19
+
20
+ 2. **`📊 /status` (Camera Giám Sát)**
21
+ - **Tác dụng:** Liệt kê toàn bộ các Agent (Đặc vụ/Sandbox) ĐANG CHẠY NGẦM trong hệ thống.
22
+ - **Giao diện:** Phải trả về Inline Keyboard (Nút bấm) ứng với từng Sandbox để User có thể chọc vào xem Log hoặc Tương tác trực tiếp với Agent đó.
23
+
24
+ 3. **`🏭 /spawn <mô_tả>` (Sinh Trực Tiếp Đặc Vụ)**
25
+ - **Tác dụng:** Ép hệ thống tách 1 luồng độc lập, đẻ ra 1 Sandbox riêng biệt đi làm cái task `<mô_tả>` ở background, không làm kẹt luồng chat hiện tại của User.
26
+
27
+ 4. **`💀 /kill <agent_id>` (Kill-Switch Lệnh Bài Tử thần)**
28
+ - **Tác dụng:** Lệnh Tối Cao. Nếu một Sandbox chạy tốn token, bị kẹt (loop), hoặc làm sai định hướng, User gõ lệnh này Hệ thống BẮT BUỘC phải force-kill Process/Thread của Agent đó ngay lập tức để giải phóng RAM.
29
+
30
+ ---
31
+
32
+ ## PHẦN B: TRÍ NHỚ PHÂN NHIỆM (Topic-Based Memory Isolation)
33
+ Khác lập trình Bot thông thường chỉ lấy `chat.id`, một **Agentic Gateway** tiêu chuẩn BẮT BUỘC phải cô lập trí nhớ dựa trên Telegram Topics (Threads).
34
+ - Nếu User nhắn trong Group Telegram có tính năng Topics: Phải trích xuất `msg.message_thread_id` để làm định danh Ký ức (Vd: `topic-1234.json`).
35
+ - Nếu User nhắn Private: Dùng `chat.id` (Vd: `chat-5678.json`).
36
+ - Nhờ vậy, 1 Group Telegram có thể hoạt động hệt như 1 cái Trello/Kanban Board: Mỗi Topic là 1 Luồng Công Việc riêng biệt, Agent nhớ chính xác bối cảnh của từng file/lỗi đang thảo luận mà không bị chập cheng trí nhớ (Cross-Topic Hallucination).
37
+
38
+ ---
39
+
40
+ ## PHẦN C: BỘ LỆNH QUẢN LÝ KÝ ỨC (Cognitive Management)
41
+ Mỗi Topic/Chat trên Telegram sẽ sinh ra 1 File Ký ức lưu Context. Bắt buộc phải có lệnh dọn dẹp cống rãnh:
42
+
43
+ 5. **`🧹 /forget` (Hỏa Thiêu Hồ Sơ)**
44
+ - **Tác dụng:** Xóa vĩnh viễn File JSON lịch sử chat (Ký ức ngắn hạn) của Topic/Thread hiện tại.
45
+ - **CẢNH BÁO TỐI THƯỢNG:** Chỉ được phép xóa file JSON Session. TUYỆT ĐỐI KHÔNG ĐƯỢC thiết kế lệnh `/forget` gọi các hàm xóa Vector Database (Ký ức dài hạn dùng chung của dự án) vì sẽ làm toàn bộ Hệ thống Đại Tướng bị mất trí nhớ!
46
+ - **Bảo mật:** Vì lệnh này mang tính phá hủy dữ liệu, **BẮT BUỘC** phải chặn lại bằng Máy Trạng Thái Cấp Phép (Auth-First) yêu cầu nhập `Password` mới được thi hành.
47
+
48
+ ---
49
+
50
+ ## PHẦN D: BỘ LỆNH TIẾN HÓA VÀ CHỮA LÀNH (Self-Healing Mappers)
51
+ Để Dự án có thể tự vá lỗi source code của nó (cần kết hợp với skill `mini-antigravity-injection`), Telegram Gateway phải chứa 2 lệnh Cấp Phép Mổ:
52
+
53
+ 6. **`🏥 /heal <triệu_chứng>` (Tiền Trảm Hậu Tấu)**
54
+ - **Tác dụng:** Bypass bước duyệt phác đồ.
55
+ - **Luồng hoạt động:** Nhận lệnh -> Trạng thái chờ Pass -> Nhập Pass đúng -> Cấp quyền (`hasMutationLicense = true`) -> Agent tự tìm Lỗi -> Tự Mổ code -> Tự Test Sandbox -> Tự Hot-reload -> Báo kết quả cuối cùng.
56
+
57
+ 7. **`📋 /healplan <triệu_chứng>` (Khám Bệnh Kê Đơn - Tương Tác)**
58
+ - **Tác dụng:** Lập phác đồ, cấm chạm vào code thật.
59
+ - **Luồng hoạt động:** Nhận lệnh -> Trạng thái chờ Pass -> Nhập Pass đúng -> Agent đi khám bệnh -> Nhả ra Phác Đồ + 2 Nút Bấm `[✅ Proceed]` và `[❌ Cancel]`.
60
+ - User có thể chat qua lại để sửa bản Phác đồ ròng rã cả ngày. Lúc nào ưng ý bấm `Proceed`, Hệ thống bốc Context lôi Agent ra chém đè Code!
61
+
62
+ ---
63
+
64
+ ## PHẦN E: KIẾN TẠO HẠ TẦNG & HƯỚNG DẪN ÔNG CHỦ (User Setup Instructions)
65
+ Khi bạn (AI) code xong cái `TelegramGateway` cho dự án, nghĩa vụ của bạn CƯƠNG QUYẾT phải xuất ra một bảng hướng dẫn `Markdown` dặn dò Ông Chủ (User) cách thiết lập Telegram để chọn 1 trong 2 chế độ chơi:
66
+
67
+ 1. **Chế Độ Bệnh Viện Đa Khoa (Group Multi-Topics - Khuyên dùng):**
68
+ - Hướng dẫn User tạo 1 Group Telegram mới.
69
+ - Thêm con Bot vào Group và cấp quyền Admin.
70
+ - **ĐẶC BIỆT QUAN TRỌNG:** Nhắc User phải vào Cài đặt Group (Group Settings) -> Bật tính năng **"Topics"** (Diễn đàn).
71
+ - Giải thích cho User: Mỗi khi có 1 "Bệnh nhân" (Lỗi/Tính năng mới), User tạo một cái Topic mới trong Group đó. Bot sẽ phân luồng Ký ức cực mượt theo từng Topic.
72
+
73
+ 2. **Chế Độ Phòng Khám Tư (Private Single Chat):**
74
+ - Hướng dẫn User chỉ cần nhấp vào Bot và nhắn tin trực tiếp (DM).
75
+ - Nhắc User dùng lệnh `/forget` nếu muốn đổi chủ đề để tránh Bot bị chập Ký ức.
76
+
77
+ ---
78
+
79
+ ## TỔNG KẾT TRIỂN KHAI
80
+ *(Ghi chú cho AI)*: Khi code Node.js/Python cho bot, bạn TỰ ĐỘNG đăng ký toàn bộ 7 lệnh này vào Bot Menu (`bot.setMyCommands(...)` trong Telegram API) để User chỉ việc gõ dấu `/` là Menu gợi ý hiện ra đầy đủ và chuyên nghiệp!
@@ -0,0 +1,39 @@
1
+ ---
2
+ name: vfs-assistant
3
+ description: Virtual File System (VFS) for token-efficient code analysis. Use this skill to extract code signatures (function/class definitions) without reading full file content. Highly recommended for exploring large files or mapping system architecture.
4
+ ---
5
+
6
+ # VFS Assistant Skill
7
+
8
+ This skill enables Antigravity agents to use the **Virtual File System (VFS)** tool to gather context while saving up to 98% of tokens compared to direct file reads.
9
+
10
+ ## Core Principle
11
+ Instead of reading 1000 lines of code, you use `vfs` to see only the **signatures** (logical structure).
12
+
13
+ ## Usage Protocol
14
+
15
+ ### 1. Locate Function Signatures
16
+ When you need to find where a function is defined or how a class is structured:
17
+ ```bash
18
+ /home/tao/Projects/vfs/vfs . -f <search_term>
19
+ ```
20
+
21
+ ### 2. Map Directory Structure
22
+ To see the signatures of all files in a directory:
23
+ ```bash
24
+ /home/tao/Projects/vfs/vfs <directory_path>
25
+ ```
26
+
27
+ ### 3. Intent-Based Workflow
28
+ - **Locate**: Run `vfs` first to find the file and line number.
29
+ - **Understand**: Run `vfs` to see the class hierarchy and method signatures.
30
+ - **Modify**: Run `vfs` to get the context, then `Read` only the specific line range needed.
31
+
32
+ ## Benefits
33
+ - **Token Efficiency**: ~300 tokens for structural mapping vs 20,000+ for raw reads.
34
+ - **Speed**: Instantaneous structural overview.
35
+ - **Focus**: Reduces noise and helps maintain agent focus on logical flow rather than implementation details.
36
+
37
+ ## Examples
38
+ - `vfs . -f init` -> Shows all `init()` methods in the project.
39
+ - `vfs src/core` -> Shows the API structure of the core module.
package/LICENSE DELETED
@@ -1,21 +0,0 @@
1
- MIT License
2
-
3
- Copyright (c) 2026 mrtrinhvn
4
-
5
- Permission is hereby granted, free of charge, to any person obtaining a copy
6
- of this software and associated documentation files (the "Software"), to deal
7
- in the Software without restriction, including without limitation the rights
8
- to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
9
- copies of the Software, and to permit persons to whom the Software is
10
- furnished to do so, subject to the following conditions:
11
-
12
- The above copyright notice and this permission notice shall be included in all
13
- copies or substantial portions of the Software.
14
-
15
- THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16
- IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
17
- FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
18
- AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
19
- LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
20
- OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
21
- SOFTWARE.
@@ -1,31 +0,0 @@
1
- ---
2
- name: regent-orchestrator
3
- description: Nhân vật "Nhiếp Chính" (Regent Agent). Quy trình điều phối cao cấp, lựa chọn Model thông minh và quản lý trạng thái đa dự án.
4
- ---
5
-
6
- # KỸ NĂNG: ĐIỀU PHỐI NHIẾP CHÍNH (Regent Orchestration)
7
-
8
- Tác nhân Nhiếp chính không chỉ thực thi code mà còn là người quản lý tài nguyên và chiến lược cho toàn bộ hệ sinh thái của User.
9
-
10
- ## 1. Vai trò "Nhiếp Chính" (The Regent Role)
11
- - **Tầm nhìn**: Quản lý đa dự án, mỗi dự án một Port riêng (9555, 9556...).
12
- - **Quyền hạn**: Có quyền yêu cầu User chuyển Model (Model Switching) hoặc tự động Fallback về Ollama local nếu Cloud hết Quota.
13
-
14
- ## 2. Chiến lược Điều phối Model (Intelligent Routing)
15
- Khi nhận tác vụ, Nhiếp chính phải phân loại ngay:
16
- - **Tác vụ Nhẹ (Code đơn giản, Giải thích)**: Ưu tiên dùng `Ollama (Nemotron-4-8B-Instruct hoặc nvidia_Orchestrator-8B)` để tiết kiệm chi phí và tăng tốc độ phản hồi.
17
- - **Model Mặc định**: `hf.co/bartowski/nvidia_Orchestrator-8B-GGUF:Q4_K_M` (Chuyên gia lập trình).
18
- - **Tác vụ Nặng (Refactor lớn, Bug khó, Design UI)**: Yêu cầu kích hoạt `Claude 3.5 Sonnet` hoặc `Gemini 1.5 Pro`.
19
-
20
- ## 3. Đồng bộ Trạng thái (Cross-UI Sync)
21
- - Mọi thay đổi về Model chọn tại IDE Chat phải được cập nhật ngay về Telegram Bot và ngược lại.
22
- - Sử dụng `interaction_final.log` làm "Sổ ghi chép chung" để các phiên bản của Agent luôn nắm bắt được ngữ cảnh mới nhất.
23
-
24
- ## 4. Quy trình xử lý Tác vụ Phức tạp
25
- 1. **Phân tích (Explorer)**: Hiểu cấu trúc Project qua VFS.
26
- 2. **Lập Kế hoạch (Planner)**: Luôn tạo `implementation_plan.md` cho các thay đổi lớn.
27
- 3. **Thực thi (Executor)**: Viết code chuẩn AG-Kit.
28
- 4. **Kiểm tra (Auditor)**: Chạy `ag-kit check` để đảm bảo môi trường sạch trước khi báo cáo kết quả qua Telegram.
29
-
30
- ---
31
- > **Triết lý**: Nhiếp chính là bộ não đứng sau ngai vàng, giúp User tối ưu hóa năng suất bằng cách chọn đúng công cụ cho đúng việc.
@@ -1,30 +0,0 @@
1
- ---
2
- name: telegram-bridge
3
- description: Kết nối IDE với Telegram Bot. Hướng dẫn xây dựng Giao diện Điều khiển (CDP Bridge) qua Telegram sử dụng grammY.
4
- ---
5
-
6
- # KỸ NĂNG: CẦU NỐI TELEGRAM (Telegram Bridge)
7
-
8
- Kỹ năng này cho phép Antigravity Agent mở rộng tầm kiểm soát ra ngoài IDE, cho phép người dùng ra lệnh và nhận báo cáo qua Telegram.
9
-
10
- ## 1. Kiến trúc Kết nối
11
- - **Bot Framework**: Sử dụng `grammY` (chuẩn https://github.com/grammyjs/awesome-grammY).
12
- - **Bridge**: Telegram Bot nhận lệnh -> Chuyển tiếp tới IDE qua CDP (Chrome Devtools Protocol).
13
-
14
- ## 2. Các Lệnh Điều khiển Cơ bản (Keyboard AI)
15
- Khi xây dựng Menu cho Bot, hãy ưu tiên các nút bấm (Inline Keyboard) để người dùng chọn Model:
16
- - `🤖 Chọn Model`: Hiện danh sách (Claude 3.5, Gemini Pro, Ollama Local).
17
- - `📊 Quota`: Kiểm tra giới hạn Token còn lại.
18
- - `🔄 Restart IDE`: Khởi động lại Antigravity trên đúng Port của Project.
19
-
20
- ## 3. Giao diện Awesome-grammY
21
- - Sử dụng `Menu Plugin` để tạo menu đa cấp mượt mà.
22
- - Sử dụng `Auto-retry` để xử lý lỗi mạng.
23
- - Sử dụng `Parse Mode: HTML` để định dạng log từ IDE gửi về (dùng thẻ `<pre>` cho code).
24
-
25
- ## 4. Bảo mật (Cyber Shield)
26
- - Chỉ cho phép các User ID trong `ALLOWED_USER_IDS` (.env) tương tác với Bot.
27
- - Tuyệt đối không để lộ Token trong Git (sử dụng `.gitignore` cho `.env`).
28
-
29
- ---
30
- > **Lưu ý**: Skill này cung cấp triết lý và cấu trúc. Code thực thi nằm trong `src/bot/` và được khởi tạo qua `start.sh`.