# Agentic Coding

> Skill family for agent-assisted software development workflows. Includes workspace isolation (git-worktrees), parallel debugging (dispatching-parallel-agents), structured multi-task development (subagent-driven-development), and lightweight code review (requesting-code-review). Install individual skills based on workflow needs.

- Skill: `jr2804/agentic-coding` (Agent Skill)
- Install (CLI): `npx skillmds@latest add jr2804/agentic-coding`
- Raw SKILL.md: https://api.skillmd.com/api/skills/jr2804/agentic-coding/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- License: MIT
- Author: jr2804 (https://skillmd.com/u/jr2804)
- Updated: 2026-09-10
- Page: https://skillmd.com/skills/jr2804/agentic-coding

---


# Agentic Coding

A modular skill family for agent-assisted software development workflows. Each skill addresses a distinct phase or pattern—install only what you need.

## Skill Family Members

| Skill | Purpose | When to Use |
| ------------------------------- | ------------------- | ----------------------------------------------------------- |
| **using-git-worktrees** | Workspace isolation | Starting feature work that needs clean branch environment |
| **dispatching-parallel-agents** | Parallel debugging | 2+ completely independent failures across different domains |
| **subagent-driven-development** | Structured workflow | Multi-task implementation with integrated two-stage review |
| **requesting-code-review** | Standalone review | Ad-hoc reviews outside structured workflows |

## Common Patterns

### Full Feature Development

```text
using-git-worktrees → subagent-driven-development
     ↓                        ↓
Create isolated          Execute plan with
feature branch           per-task subagents
                         + integrated review
```

### Parallel Bug Investigation

```text
dispatching-parallel-agents
     ↓
Multiple subagents
investigate unrelated
failures simultaneously
```

### Quick Post-Task Review

```text
requesting-code-review
     ↓
Lightweight review
after single task
or before merge
```

## Installation

Install individual skills based on your workflow:

```bash
# Need workspace isolation?
saddle-cli install agentic-coding/using-git-worktrees

# Need parallel debugging?
saddle-cli install agentic-coding/dispatching-parallel-agents

# Need structured development workflow?
saddle-cli install agentic-coding/subagent-driven-development

# Need standalone code review?
saddle-cli install agentic-coding/requesting-code-review
```

## Relationship to Other Skills

- **coding-discipline**: Apply behavioral discipline within any agentic workflow
- **python-ultimate**: Technical standards for Python implementation
- **output-quality**: Review final output for anti-patterns

## Decision Flowchart

```dot
digraph which_skill {
    "Starting feature work?" [shape=diamond];
    "Need clean workspace?" [shape=diamond];
    "using-git-worktrees" [shape=box];
    "Multiple failures?" [shape=diamond];
    "Independent domains?" [shape=diamond];
    "dispatching-parallel-agents" [shape=box];
    "Single failure path" [shape=box];
    "Have implementation plan?" [shape=diamond];
    "Multi-task plan?" [shape=diamond];
    "subagent-driven-development" [shape=box];
    "requesting-code-review" [shape=box];

    "Starting feature work?" -> "Need clean workspace?" [label="yes"];
    "Starting feature work?" -> "Have implementation plan?" [label="no"];
    "Need clean workspace?" -> "using-git-worktrees" [label="yes"];
    "Need clean workspace?" -> "Have implementation plan?" [label="no"];
    "Have implementation plan?" -> "Multi-task plan?" [label="yes"];
    "Have implementation plan?" -> "Multiple failures?" [label="no"];
    "Multi-task plan?" -> "subagent-driven-development" [label="yes"];
    "Multi-task plan?" -> "requesting-code-review" [label="no"];
    "Multiple failures?" -> "Independent domains?" [label="yes"];
    "Multiple failures?" -> "Single failure path" [label="no"];
    "Independent domains?" -> "dispatching-parallel-agents" [label="yes"];
    "Independent domains?" -> "Single failure path" [label="no"];
}
```