Where to Find the Full Ponytail Skill Definition: Complete File Guide

The complete Ponytail skill definition resides in skills/ponytail/SKILL.md, which contains the command templates, prompt rules, and usage examples that agents invoke via /ponytail commands.

The DietrichGebert/ponytail repository provides a structured plugin architecture for LLM agents. Understanding exactly where the Ponytail skill definition lives is essential for customizing behavior, debugging agent responses, or manually registering the skill in compatible environments.

Primary Location of the Ponytail Skill Definition

The authoritative source of truth for Ponytail's capabilities is located at skills/ponytail/SKILL.md. This markdown file houses the complete command-prompt template, textual injection rules, and usage examples that govern how agents interpret the /ponytail interface.

Inside the SKILL.md File Structure

According to the repository source code, SKILL.md defines the "ponytail" skill that agents recognize and invoke. When you issue commands such as /ponytail or ponytail:<sub-skill>, the agent references the definitions stored in this specific file to determine available actions, parameter formats, and response constraints.

How the Skill Definition Is Registered

The mechanism that exposes the skill to agents is implemented in __init__.py. This initialization file automatically discovers skill definitions within the skills/ directory hierarchy and registers them with the host context.

The registration uses the context's register_skill method:

ctx.register_skill("ponytail", Path("skills/ponytail/SKILL.md"))

This call binds the string identifier "ponytail" to the markdown definition file, enabling the plugin to inject the skill's prompt rules into every LLM turn.

Alternative Entry Points and Hooks

Beyond the primary SKILL.md file, the repository provides additional access patterns for different agent architectures and initialization scenarios.

Lifecycle Hooks for Automatic Loading

For agents that support automated hook loading, the hooks/ directory contains configuration files such as hooks/qoder-hooks.json. These hooks automatically load the Ponytail skill definition during agent initialization without requiring explicit manual registration calls in the host code.

AGENTS.md Fallback Instructions

The repository includes AGENTS.md as a fallback instruction set. This file provides a condensed, static version of the skill rules for host environments that only read textual rule files rather than registering dynamic skills through the plugin interface.

Programmatic Access to the Skill Definition

You can load and inspect the skill definition directly using standard Python file operations. The following pattern demonstrates how the plugin internally accesses the markdown content:

from pathlib import Path
ROOT = Path(__file__).parent.parent
skill_md = (ROOT / "skills" / "ponytail" / "SKILL.md").read_text(encoding="utf-8")
print(skill_md)   # prints the full skill definition

When working within an LLM-enabled environment such as Claude Code, you activate specific skill modes using the command syntax defined in SKILL.md:

/ponytail ultra      # activates the "ultra" mode using the definition in SKILL.md

Summary

• The complete Ponytail skill definition lives in skills/ponytail/SKILL.md
• Registration occurs in __init__.py via ctx.register_skill("ponytail", Path(...))
• Lifecycle hooks in hooks/ enable automatic loading for compatible agents
• AGENTS.md serves as a static fallback for rule-only host environments

Frequently Asked Questions

What file contains the complete Ponytail skill definition?

The file skills/ponytail/SKILL.md contains the complete definition, including command templates, prompt rules, and usage examples that agents use when processing /ponytail commands.

How do I register the Ponytail skill manually in my agent?

Import the skill file path and call ctx.register_skill("ponytail", Path("skills/ponytail/SKILL.md")) as implemented in __init__.py. This binds the skill identifier to the markdown definition for the host context.

What is the difference between SKILL.md and AGENTS.md?

SKILL.md contains the full dynamic skill definition used for plugin registration and runtime prompt injection, while AGENTS.md provides a static fallback version for hosts that consume rules as plain text without supporting dynamic skill loading.

Where are the lifecycle hooks for automatic skill loading located?

Lifecycle hooks are stored in the hooks/ directory, with files such as qoder-hooks.json configuring automatic skill initialization for agents that support hook-based plugin discovery.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}