Skill Command

Generate a new Skill with standard structure and boilerplate.

Usage

/skill <category>/<skill-name>

Example

/skill input/calendar-input
/skill process/garden-writer
/skill output/git-output
/skill meta/skill-builder

What This Does

1. Creates skill directory at

   skills/<skill-name>/

2. Generates skill.yaml with category-appropriate defaults
3. Generates skill.py with

   execute(context)

   boilerplate
4. Creates test file at

   bsage/tests/test_skill_<name>.py

Generated Structure

For

/skill input/calendar-input

:

skills/
└── calendar-input/
    ├── skill.yaml      # Metadata + trigger config
    └── skill.py         # execute(context) function

bsage/
└── tests/
    └── test_skill_calendar_input.py

Generated Files

skills/calendar-input/skill.yaml

name: calendar-input
version: 1.0.0
author: bslab
category: input
is_dangerous: false
entrypoint: skill.py::execute
trigger:
  type: cron
  schedule: "*/15 * * * *"
rules:
  - garden-writer
description: Google Calendar 일정을 2nd Brain으로 수집

skills/calendar-input/skill.py

from __future__ import annotations


async def execute(context):
    """Google Calendar 일정을 수집하여 seed로 저장."""
    context.logger.info("calendar_input_start")

    creds = await context.credentials.get("google-calendar")
    # Skill 내부에서 직접 API 연결 처리
    events = await fetch_calendar_events(creds)

    await context.garden.write_seed("calendar", {"events": events})

    context.logger.info("calendar_input_complete", event_count=len(events))
    return {"collected": len(events)}

bsage/tests/test_skill_calendar_input.py

import pytest
from unittest.mock import AsyncMock, MagicMock


@pytest.fixture
def mock_context():
    context = MagicMock()
    context.logger = MagicMock()
    context.credentials = MagicMock()
    context.credentials.get = AsyncMock(return_value={"client_id": "test"})
    context.garden = AsyncMock()
    context.garden.write_seed = AsyncMock()
    return context


@pytest.mark.asyncio
async def test_execute_collects_events(mock_context):
    from importlib.util import spec_from_file_location, module_from_spec
    spec = spec_from_file_location("skill", "skills/calendar-input/skill.py")
    module = module_from_spec(spec)
    spec.loader.exec_module(module)

    result = await module.execute(mock_context)

    assert result["collected"] >= 0
    mock_context.garden.write_seed.assert_called_once()

Category Templates

InputSkill

category: input
trigger:
  type: cron
  schedule: "*/15 * * * *"
rules:
  - garden-writer

ProcessSkill

category: process
is_dangerous: false    # true if external side effects

OutputSkill

category: output

MetaSkill

category: meta
is_dangerous: false

Next Steps

After generation:

1. Fill in skill.py implementation
2. Update skill.yaml fields (rules, trigger)
3. Add test cases for error paths
4. Run tests with

   /test

5. Verify Skill loads correctly with SkillLoader