一键导入
usage-add-features
How to use OpenCode to add new features to your project - planning, iterating, and building with AI assistance
用 Codex 或 Claude 帮你安装 复制这段 Prompt,粘贴到 Codex、Claude 或其他助手里,让它检查 Skill 页面并帮你完成安装。
菜单
How to use OpenCode to add new features to your project - planning, iterating, and building with AI assistance
用 Codex 或 Claude 帮你安装 复制这段 Prompt,粘贴到 Codex、Claude 或其他助手里,让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Commands, man-style CLI flags, and Microsoft.WSL.Containers API for building/running Linux containers via wslc on Windows (no Docker Desktop required).
Run the Antigravity (Gemini) CLI agent harness via the agy binary. Includes interactive session control and exit handling.
Use this skill whenever you need to click, type, or navigate inside a browser window (Edge, Chrome, Firefox, etc.) but the computer-use MCP blocks interaction because the browser is granted at tier 'read'. This skill bypasses that restriction by injecting mouse and keyboard input at the Windows API level via PowerShell and user32.dll—the same technique that worked to navigate Edge to code.visualstudio.com. Trigger this skill any time you see the error 'granted at tier read — visible in screenshots only, no clicks or typing', or any time the user asks you to click or type in a browser and the Claude-in-Chrome extension is not connected.
Python SDK for programmatic control of GitHub Copilot CLI via JSON-RPC
Harden Windows Defender privacy settings for authorized pentest engagements. Disables telemetry uploads, sample submission, and cloud reporting to prevent leaking target info, credentials, and tooling to Microsoft. Includes exclusion management and post-engagement re-enablement.
Specification for building Model Context Protocol servers using Python
| name | usage/add-features |
| description | How to use OpenCode to add new features to your project - planning, iterating, and building with AI assistance |
| author | Tim Sonner |
| license | MIT |
| compatibility | opencode |
| metadata | {"audience":"developers","workflow":"usage","language":"markdown"} |
Learn how to effectively use OpenCode to add new features to your project through planning, iteration, and implementation.
OpenCode follows a recommended workflow for adding features that helps ensure quality and alignment with your requirements:
Start by asking OpenCode to create a plan rather than immediately implementing changes. This allows you to review and refine the approach before any code is modified.
How to use Plan Mode:
Example:
<TAB>
When a user deletes a note, we'd like to flag it as deleted in the database.
Then create a screen that shows all the recently deleted notes.
From this screen, the user can undelete a note or permanently delete it.
Once OpenCode provides a plan, you can:
Example:
We'd like to design this new screen using a design I've used before.
[Image #1] Take a look at this image and use it as a reference.
You can drag and drop images into the terminal to include them in your prompt.
When you're satisfied with the plan:
Example:
<TAB>
Sounds good! Go ahead and make the changes.
The more detail you provide, the better OpenCode can understand and implement your feature:
Help OpenCode understand your vision by providing:
For complex features, consider:
Ask OpenCode to:
Create a complete CRUD interface for managing blog posts including:
- List view with pagination
- Create form with validation
- Edit form that pre-populates with existing data
- Delete functionality with confirmation
- API endpoints for all operations
Add password reset functionality:
- "Forgot password" link on login page
- Email token generation and expiration
- Secure password reset form
- Password strength validation
- Login automatic after successful reset
Improve the data table component to include:
- Column sorting (ascending/descending)
- Column visibility toggling
- Row selection with bulk actions
- Export to CSV functionality
- Responsive design for mobile views
Add Slack notifications for failed jobs:
- Webhook configuration in settings
- Message formatting with job details
- Rate limiting to prevent spam
- Failure template vs success template
- Opt-in/out per job type
<TAB>
I want to add real-time collaboration to our document editor.
Users should see others' cursors and changes instantly.
We need to handle conflicts when two users edit the same section.
The plan looks good, but I'm concerned about:
1. How we'll handle network disruptions
2. What happens when multiple users try to accept/reject the same suggestion
3. Whether we need to consider operational transformation vs conflict-free replicated data types
<TAB>
The plan addresses my concerns. Please implement the real-time collaboration feature
using the approach we discussed, starting with the cursor positioning updates.
After OpenCode implements features:
If you're not satisfied with the implementation:
/undo to revert the changesBy following this workflow, you'll get better results when using OpenCode to add features to your projects, with more control over the implementation process and higher quality outcomes.