Agent Asset Studio

# Naming Convention

## 1. Domain Prefixing

- **Purpose**: Categorization/tagging (since folders are not always nested).
- **Rule**: Place 1-2 domain tokens at the front.
- **Sub-skill Rule**: If nested under a parent skill, omit domain tokens already clear from the parent skill name.
- **Example**: `openspec-apply-change.md`, `agent-asset-studio.md`

## 2. Naming Types by Skill Complexity

| Type | Target Skill | Format | Example |
| --- | --- | --- | --- |
| **Verb** | Simple skill (single action) | `<domain>-<verb>-<details>.md` | `coder-generate-code.md` |
| **Object** | Complex skill (multi-action object) | `<domain>-<object>.md` | `task-manager.md` |
| **Place** | Routing/Hub skill (consolidates capabilities) | `<domain>-<details>-<place>.md` | `reviewer-console.md` |

*Note: `<place>` must be a location-based word (e.g., `studio`, `console`, `hub`, `portal`, `workspace`).*


# Routing Skill Structure

Terse reference for designing and structuring routing skills.

## 1. Purpose

- **Efficient Indexing**: Resolves the inefficiency of scanning numerous separate skill directories as capabilities expand.
- **Cost Optimization**: Consolidates common scripts and assets under one master router, minimizing context size and inference overhead.

## 2. Core Components

### `SKILL.md` (The Router)

- Acts as the primary entry point.
- Does NOT contain the actual execution workflows for individual tasks.
- Guides the agent to read `INDEX.csv` and route the user's request to the appropriate sub-skill.

### `INDEX.csv` (The Registry)

- A lightweight index containing minimal data needed for the LLM to make an accurate routing decision.
- Format: `name,overview,keywords,trigger,exclusion`
- The agent reads this first to locate the target sub-skill without loading all sub-skill instructions into context.

## 3. Sub-Skill Organization

Routing skills allow two layouts for their children within the `sub-skills/` directory:

1. **Flat Markdown** (For simple tasks)
   - Path: `sub-skills/<skill-name>.md`
   - Ideal for concise, single-file instructions.

2. **Nested Directory** (For complex tasks or migrated skills)
   - Path: `sub-skills/<skill-name>/SKILL.md`
   - Used when migrating an existing skill to preserve its isolated assets, references, or folder structure.


from . import agent, rule, skill

__all__ = [
    "skill",
    "rule",
    "agent",
]


from typing import Final

from core.schema import FrontmatterSchema

from .common import (
    ALLOWED_RESOURCES,
    COMMON_DESCRIPTION_FIELD,
    COMMON_NAME_FIELD,
    DEFAULT_AGENT_DESCRIPTION,
)

DEFAULT_DESCRIPTION: Final[str] = DEFAULT_AGENT_DESCRIPTION

# Configuration for Agent Validation
FRONTMATTER_SCHEMA: Final[FrontmatterSchema] = FrontmatterSchema([
    COMMON_NAME_FIELD,
    COMMON_DESCRIPTION_FIELD,
])


import re
from typing import Final

from core.schema import FrontmatterField

ALLOWED_RESOURCES: Final[set[str]] = {
    "scripts",
    "references",
    "assets",
    "prompts",
    "configs",
    "sub-skills",
}

NAME_PATTERN: Final[re.Pattern[str]] = re.compile(r"^[a-z0-9]+(?:-[a-z0-9]+)*$")

COMMON_NAME_FIELD: Final[FrontmatterField] = FrontmatterField(
    "name",
    is_required=True,
    max_length=64,
    regex_pattern=r"^[a-z0-9]+(?:-[a-z0-9]+)*$",
    regex_message="name은 소문자 영문, 숫자, 단일 하이픈만 사용할 수 있습니다.",
)

COMMON_DESCRIPTION_FIELD: Final[FrontmatterField] = FrontmatterField(
    "description",
    is_required=True,
    max_length=1024,
)

DEFAULT_AGENT_DESCRIPTION: Final[str] = (
    "TODO: Describe the persona, goal, and responsibilities of this agent. Use to "
    "instantiate specialized sub-agents or define main orchestrators."
)

DEFAULT_RULE_DESCRIPTION: Final[str] = (
    "TODO: Describe what this rule restricts, enforces, or guides. Use when the agent "
    "needs to adhere to specific protocols, coding styles, or behavior limits."
)

DEFAULT_SKILL_DESCRIPTION: Final[str] = (
    "TODO: Describe what this skill does and when to use it. Use when the user