Files
opencode-skill/skills/skill-creator/SKILL.md
2026-03-08 23:03:19 +07:00

197 lines
4.8 KiB
Markdown

---
name: skill-creator
description: Create new OpenCode skills with proper structure, SKILL.md format, and script templates. Use this skill when you need to create a new OpenCode skill.
---
# Skill Creator
Guide and tools for creating new OpenCode skills.
## Quick Start
```bash
python3 scripts/create_skill.py <skill-name> "<description>"
```
## SKILL.md Format (Required)
Every skill must have a `SKILL.md` file with YAML frontmatter:
```yaml
---
name: skill-name
description: Brief description. Use when user wants to [specific action].
---
# Skill Name
Brief explanation of what this skill does.
## Commands
| Command | Args | Description |
|---------|------|-------------|
| `command1` | `<arg>` | What it does |
## Options
| Option | Default | Range | Description |
|--------|---------|-------|-------------|
| `--option` | 100 | 1-1000 | What it does |
## Examples
```bash
python3 scripts/script.py command "arg" --option 50
```
## Output Format
- Success: `Result: filename [id]`
- Error: `Error: message` (to stderr)
## Notes
- Required environment variables
- Important constraints
```
## Frontmatter Rules
| Field | Required | Rules |
|-------|----------|-------|
| `name` | Yes | 1-64 chars, lowercase alphanumeric + hyphens, no leading/trailing/consecutive hyphens |
| `description` | Yes | 1-1024 chars, specific enough for agent to choose correctly |
| `license` | No | e.g., MIT |
| `compatibility` | No | e.g., opencode |
| `metadata` | No | String-to-string map |
## Directory Structure
```
skills/
└── skill-name/
├── SKILL.md # Required: skill definition
└── scripts/
├── main_script.py # Executable script
├── .env.example # Required: env var template
└── requirements.txt # Optional: Python deps
```
## Script Best Practices
### 1. Load Environment Variables
```python
def load_env():
env_path = Path(__file__).parent / ".env"
if env_path.exists():
for line in env_path.read_text().splitlines():
line = line.strip()
if line and not line.startswith("#") and "=" in line:
k, v = line.split("=", 1)
os.environ.setdefault(k.strip(), v.strip().strip("\"'"))
load_env()
API_TOKEN = os.environ.get("API_TOKEN")
```
### 2. Handle API Responses (Binary + JSON)
APIs may return raw binary or JSON with base64. Handle both:
```python
response = requests.post(url, headers=headers, json=payload, timeout=300)
response.raise_for_status()
content_type = response.headers.get("Content-Type", "")
if "image/" in content_type or "application/octet-stream" in content_type:
# Raw binary response
data = response.content
else:
# JSON with base64
result = response.json()
if isinstance(result, list) and len(result) > 0:
image_data = result[0].get("data", "")
if image_data.startswith("data:"):
data = base64.b64decode(image_data.split(",", 1)[1])
else:
data = base64.b64decode(image_data)
```
### 3. Send Base64 (Plain, Not Data URI)
Some APIs expect plain base64, not data URI:
```python
import base64
with open(image_path, "rb") as f:
image_bytes = f.read()
# Plain base64 (no data: prefix)
b64_string = base64.b64encode(image_bytes).decode("utf-8")
```
### 4. Output Format
Follow OpenCode conventions:
```python
# Success with ID
print(f"Result: {filename} [{timestamp}]")
# Error to stderr
print(f"Error: {message}", file=sys.stderr)
sys.exit(1)
```
### 5. CLI Arguments
Use argparse for clean CLI:
```python
parser = argparse.ArgumentParser(description="What this does")
parser.add_argument("required_arg", help="Description")
parser.add_argument("--optional", type=int, default=100, help="Description")
args = parser.parse_args()
```
## .env.example Template
```
# API credentials
# Get your token from https://service.com/account
#
# WARNING: Never commit actual credentials!
API_TOKEN=your_api_token_here
```
## Installation Paths
| Type | Path |
|------|------|
| Global | `~/.config/opencode/skills/<name>/SKILL.md` |
| Project | `./.opencode/skills/<name>/SKILL.md` |
## Common Issues
| Issue | Solution |
|-------|----------|
| 400 Bad Request | Check payload format - may need flat JSON, not nested |
| Skill not found | Verify path is `skills/<name>/SKILL.md` (plural "skills") |
| API token not loaded | Check .env is in same directory as script |
| Binary response fails | Check Content-Type header, handle raw bytes |
## Checklist for New Skills
- [ ] `SKILL.md` with required frontmatter (name, description)
- [ ] `scripts/` directory with main script
- [ ] `scripts/.env.example` with placeholder credentials
- [ ] `scripts/requirements.txt` if external deps needed
- [ ] Script handles both binary and JSON responses
- [ ] Output follows format: `Result: name [id]`
- [ ] Errors go to stderr with `sys.exit(1)`