Skip to content

About

Cross-harness design patterns and ready-to-use skills for AI coding agents (Claude Code, Cursor, Windsurf, Copilot, AGENTS.md), with a zero-dependency sync CLI.

Resources

Stars

0 stars

Watchers

0 watching

Forks

Repository files navigation

Agentic Coding Design Patterns & Skills

License: MIT Zero Dependencies Harnesses Supported

An open, cross-harness repository of best practices, architectural design patterns, and executable skills for AI coding agents. Grounded in peer-reviewed literature and production-grade architectures from leading AI engineering labs (Anthropic, Google DeepMind, AWS Prescriptive Guidance, Princeton, UC Berkeley).

Designed to be canonical, portable, and ready to sync into any agent harness.


🌟 Supported Agent Harnesses

Harness Configuration Files Generated Skill / Command Format
Claude Code .claude/commands/*.md, CLAUDE.md Slash commands with $ARGUMENTS interpolation
Antigravity (AGY) .agents/skills/<name>/SKILL.md, GEMINI.md Native skill bundles with YAML frontmatter
Cursor .cursor/rules/*.mdc, .cursorrules MDC rule system with glob filters
Roo Code & Cline .clinerules, .roomodes Custom modes and operational instructions
Windsurf .windsurfrules Cascade contextual memory rules
GitHub Copilot .github/copilot-instructions.md Standard instructions guide
Universal / Aider AGENTS.md Repository-wide agent conventions

🧭 The 15 Core Agentic Patterns & Commands

flowchart LR
    subgraph Discovery["1. Discovery & Design"]
        direction TB
        G["/grill-me<br/><i>(Socratic Elicitation)</i>"]
        GD["/grill-me-with-docs<br/><i>(Grounded Inquiry)</i>"]
        S["/to-spec<br/><i>(RFC 2119 PRD)</i>"]
        T["/to-tickets<br/><i>(DAG Decomposition)</i>"]
        CW["/clear-writing<br/><i>(Empirical Clarity)</i>"]
        TE["/teach<br/><i>(Pedagogical Workspace)</i>"]
    end

    subgraph Execution["2. Execution & Quality"]
        direction TB
        I["/implement<br/><i>(TDD Red-Green)</i>"]
        SC["/self-critic<br/><i>(Verbal Reflexion)</i>"]
        CR["/code-review<br/><i>(Adversarial Audit)</i>"]
    end

    subgraph Meta["3. Autonomous & Meta Loops"]
        direction TB
        GL["/goal<br/><i>(Autonomous ReAct)</i>"]
        B["/boost<br/><i>(Tree-of-Thoughts)</i>"]
        TW["/teamwork<br/><i>(Subagent Swarm)</i>"]
        SL["/schedule-loop<br/><i>(Reconciliation Loop)</i>"]
        RL["/retro-learn<br/><i>(Rule Distillation)</i>"]
        V["/visualize<br/><i>(Mermaid Modeling)</i>"]
    end

    Discovery --> Execution --> Meta
Loading
Command Category Purpose Theoretical & Industry Basis
/clear-writing Communication Fluff-free technical explanations & builder writeups (ASD-STE100 + Karpathy style). ASD-STE100 Issue 9; Karpathy Builder Ethos
/grill-me Discovery Interactive requirements interview; walks down the design tree resolving decisions one-by-one. Socratic Method; Domain-Driven Design (Evans)
/grill-me-with-docs Discovery Pre-retrieves official library documentation before asking technical architecture questions. Retrieval-Augmented Generation (Lewis et al.)
/to-spec Discovery Converts requirements and transcripts into an RFC-grade specification with strict boundaries. RFC 2119 (Bradner); Amazon PR/FAQ; ADR (Nygard)
/to-tickets Discovery Breaks specifications into an ordered Directed Acyclic Graph (DAG) of atomic, verifiable tickets. Work Breakdown Structure; BDD (North); Kahn DAG
/teach Learning Stateful multi-session learning workspace with retrieval practice, ZPD, and ADR-style learning records. Desirable Difficulties (Bjork); ZPD (Vygotsky); AI Hero (Pocock)
/implement Execution Executes tickets using Test-Driven Development (Red-Green-Refactor) and atomic git commits. Test-Driven Development (Beck); Fowler Refactoring
/self-critic Quality Pre-execution internal audit checking for hallucinations, broken invariants, and boundary conditions. Reflexion (Shinn et al., NeurIPS 2023); Self-Refine
/code-review Quality Senior staff adversarial code review evaluating correctness, security, concurrency, and performance. Fagan Inspection; Google Code Review Practices; OWASP
/retro-learn Memory Distills post-session corrections and error recoveries into permanent, concise harness rules. Voyager Skill Library (Wang et al.); MemGPT (Packer)
/goal Autonomy Autonomous long-running execution loop bounded by an empirical verification oracle and step budget. ReAct (Yao et al., ICLR 2023); AWS Agentic Loops
/boost Deliberation Multi-perspective debate (Architect, Security Auditor, SRE, Hacker) using Tree-of-Thoughts. Tree of Thoughts (Yao et al., NeurIPS 2023)
/teamwork Multi-Agent Coordinates isolated subagents (Researcher, Implementer, Reviewer) to prevent context contamination. Actor Model (Hewitt); AWS Coordinator-Subagent Hierarchy
/schedule-loop Maintenance Idempotent recurring loop for health checks, drift remediation, and dependency monitoring. Kubernetes Reconciliation Loop (Burns)
/visualize Modeling Generates standard Mermaid diagrams (C4 component, sequence, state machine, ER diagrams). C4 Model (Brown); Mermaid Specification

🚀 Quick Start & Installation

Option 1: Install into Any Target Repository (Zero Install)

You can sync these patterns directly into any existing repository using Python (standard library only, zero external packages):

# Sync into current project for all harnesses
python3 /path/to/agentic-coding-design-patterns/cli/sync.py

# Or sync to specific harnesses (e.g. Claude Code and Cursor only)
python3 /path/to/agentic-coding-design-patterns/cli/sync.py --target claude,cursor --output-dir /path/to/your/project

# Or use the bash wrapper
./install.sh --target antigravity,claude --output-dir ~/projects/my-app

Option 2: List Available Patterns

python3 cli/sync.py --list

📂 Repository Layout

agentic-coding-design-patterns/
├── skills/                     # Canonical single source of truth for all skills
│   ├── grill-me/SKILL.md
│   ├── grill-me-with-docs/SKILL.md
│   ├── to-spec/SKILL.md
│   ├── to-tickets/SKILL.md
│   ├── implement/SKILL.md
│   ├── code-review/SKILL.md
│   ├── retro-learn/SKILL.md
│   ├── goal/SKILL.md
│   ├── boost/SKILL.md
│   ├── teamwork/SKILL.md
│   ├── schedule-loop/SKILL.md
│   ├── self-critic/SKILL.md
│   └── visualize/SKILL.md
├── templates/                  # Reusable engineering templates
│   ├── prd_spec_template.md    # RFC 2119 PRD / Spec template
│   ├── ticket_template.md      # DAG task decomposition template
│   ├── code_review_template.md # Multi-lens review rubric
│   └── adr_template.md         # Architecture Decision Record
├── docs/                       # In-depth architectural documentation & citations
│   ├── agentic_design_patterns.md
│   └── harness_matrix.md
├── cli/                        # Zero-dependency Python sync compiler
│   ├── sync.py
│   └── __main__.py
├── install.sh                  # Shell installer wrapper
└── README.md

📚 Academic & Industry Citations

  • Yao, S., et al. (2022). ReAct: Synergizing Reasoning and Acting in Language Models. ICLR 2023.
  • Shinn, N., et al. (2023). Reflexion: Language Agents with Verbal Reinforcement Learning. NeurIPS 2023.
  • Yao, S., et al. (2023). Tree of Thoughts: Deliberate Problem Solving with Large Language Models. NeurIPS 2023.
  • Wang, G., et al. (2023). Voyager: An Open-Ended Embodied Agent with Large Language Models. arXiv:2305.16291.
  • Packer, C., et al. (2023). MemGPT: Towards LLMs as Operating Systems. arXiv:2310.08560.
  • Amazon Web Services (2024). Agentic Design Patterns - Prescriptive Guidance. AWS Architecture Series.
  • Huang, J. (2024). Designing AI Agents. Manning Publications.
  • Beck, K. (2002). Test-Driven Development: By Example. Addison-Wesley.
  • Fagan, M. E. (1976). Design and code inspections to reduce errors in program development. IBM Systems Journal.
  • Bradner, S. (1997). Key words for use in RFCs to Indicate Requirement Levels. RFC 2119.

📄 License

Released under the MIT License.

About

Cross-harness design patterns and ready-to-use skills for AI coding agents (Claude Code, Cursor, Windsurf, Copilot, AGENTS.md), with a zero-dependency sync CLI.

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages