# RAW Agentic Coding: ZERO to Agent SKILL Source: https://www.youtube.com/watch?v=X2ciJedw2vU Recap page: https://rapidrecap.app/video/X2ciJedw2vU Generated: 2025-12-08T14:35:28.304+00:00 --- ## Quick Overview The creator builds a custom, reusable agent skill named "Fork Terminal Skill" from scratch using a structured Markdown format that defines purpose, variables, instructions, workflow, and cookbooks for specific LLMs (Claude, Codex, Gemini) to enable agents to fork new terminal windows for specific tasks like running commands or summarizing codebases, demonstrating its capability by successfully executing a `clays` command and then demonstrating context-aware summarization for different models. **Key Points:** - The core concept demonstrated is building an agent skill from scratch using a structured Markdown file, where the skill is designed to fork new terminal sessions for agents. - The skill documentation outlines specific instructions for Claude, Codex CLI, and Gemini CLI agents, including model preferences (e.g., Claude: Haiku/Sonnet, Codex: gpt-3.5-codex-mini, Gemini: gemini-2.5-flash) and required flags like `--dangerously-skip-permissions` or `--yolo`. - The workflow section explicitly details the steps: Read the primary tool file (`fork_terminal.py`), follow the Cookbook to determine the tool, and execute the tool. - The agent successfully executed an `ffmpeg --help` command in a forked terminal, demonstrating its ability to run raw CLI commands in a new context. - When prompted to summarize the codebase, the agent correctly identified and utilized the `all_skills.md` and `prime.md` files, showing it understood the environment structure. - The agent successfully handled context-aware summarization by using the `fork_summary_user_prompt.md` file, which includes prior conversation history, proving its ability to maintain context across forks. - The final structure of the skill is highly modular, separating concerns into directories like `commands`, `skills`, `cookbook`, `prompts`, and `tools`. ![Screenshot at 00:37: The video illustrates the core concept of agent skill development using a Venn diagram showing that effective skills require the intersection of CONTEXT, TOOLS, MODEL, and PROMPT, emphasizing that the skill design must account for all four components.](https://ss.rapidrecap.app/screens/X2ciJedw2vU/00-00-37.png) **Context:** The video is a technical tutorial by IndyDevDan demonstrating the process of creating a custom, reusable agent skill within an agentic coding framework, likely related to the Cursor or similar AI-powered IDE environments. The specific skill being built is named "Fork Terminal Skill," designed to allow an agent to spawn new terminal windows to execute commands or access documentation relevant to the current task, which requires complex orchestration between different LLMs like Claude, Codex, and Gemini. ## Detailed Analysis The creator details the process of building a complex, reusable agent skill called "Fork Terminal Skill," designed to manage terminal operations across various AI models (Claude, Codex, Gemini). The skill is defined using a structured Markdown file that outlines its Purpose, Variables (including model preferences like `DEFAULT_MODEL: sonnet`, `FAST_MODEL: haiku`, and specific model settings for Codex and Gemini CLIs), Instructions, a detailed Workflow, and a Cookbook section detailing how to route requests for different agents (Claude Code, Codex CLI, Gemini CLI) using specific cookbook files. The creator demonstrates the skill's functionality by successfully executing commands like `clays` (which forks a new terminal session for Claude to run `ffmpeg --help`) and then demonstrating context-aware operation by using the agent to summarize the codebase, feeding it previous conversation history via the `fork_summary_user_prompt.md` file. The successful execution proves the skill can correctly read documentation, identify necessary tools, and fork new, correctly configured terminal environments for different underlying models, showcasing a robust, modular approach to agent development. ### Chapter 1 - Skills From Scratch: Introduction to agentic coding and the need to step back from complex setups to focus on fundamental skill creation. - The concept of Agent Skill is introduced as a combination of Prompts + Code + Consistency + Reusability. - The goal is to build a 'Fork Terminal Skill' from scratch. ### Chapter 2 - Begin with the End: Emphasizes defining the desired outcome (deliverable) before starting development, contrasting this with starting without a clear goal. ### Chapter 3 - Initialize Tooling: Details the initial project structure setup, including directories for `commands`, `skills`, `cookbook`, `prompts`, and `tools`, and the initial use of `git init`. ### Chapter 4 - Agent.md Files: Focuses on creating the instruction files for different agents (Claude, Codex, Gemini) within the `cookbook` directory, specifying model parameters (e.g., Haiku, gpt-3.5-codex-max) and instructions for handling raw CLI commands. ### Chapter 5 - Agent Forks: Demonstrates the core functionality by executing commands that fork new terminal sessions using different agents (Claude, Codex, Gemini) to perform tasks like summarizing documentation or running CLI help commands, showing that the skill correctly routes tasks based on the prompt structure and agent configuration. ![Screenshot at 00:01: IndyDevDan introducing the video and the context of agent skills.](https://ss.rapidrecap.app/screens/X2ciJedw2vU/00-00-01.png) ![Screenshot at 00:14: The formula for skill development: Skills = Prompts + Code + Consistency + Reusability, culminating in the title 'AGENT SKILL'.](https://ss.rapidrecap.app/screens/X2ciJedw2vU/00-00-14.png) ![Screenshot at 00:46: Visual representation of the skill structure being built, showing folders like .claude, skills, cookbook, prompts, and tools.](https://ss.rapidrecap.app/screens/X2ciJedw2vU/00-00-46.png) ![Screenshot at 02:20: The creator starts sketching out the plan in a notebook, emphasizing the 'Begin with the End' principle.](https://ss.rapidrecap.app/screens/X2ciJedw2vU/00-02-20.png) ![Screenshot at 03:34: The resulting file structure for the 'fork-terminal-skill' project is displayed, including the necessary SKILL.md and tool files.](https://ss.rapidrecap.app/screens/X2ciJedw2vU/00-03-34.png)