The Deska blog
Using Agents Only for Boilerplate: A Valid Strategy
Explore why using agents only for boilerplate is a valid strategy for developers who want to maintain control while accelerating repetitive setup tasks.
· 10 min read
The rise of autonomous AI assistants has led to a polarization in development workflows. Some engineers push for full autonomy, while others remain skeptical of the architectural quality produced by large language models. A middle ground has emerged where developers are using agents only for boilerplate to bypass the most tedious parts of a project without surrendering creative control over the core logic. This strategy allows you to maintain high standards for your primary business logic while leveraging the speed of AI for repetitive setup tasks, configuration files, and standard API structures.
The Case for Restricted Agent Usage
Adopting AI in software development does not have to be an all or nothing proposition. When you use agents only for boilerplate, you are essentially treating the AI as an advanced scaffolding engine rather than a lead architect. This approach mitigates many of the risks associated with AI generated code, such as hallucinated dependencies or subtle logic bugs that are difficult to debug in complex systems.
Boilerplate code is predictable, follows established patterns, and is often required for modern frameworks to function. Because these patterns are well documented in training data, agents tend to be highly accurate in this specific domain. By delegating these tasks, you save hours of manual typing and documentation lookup.
Reducing Cognitive Load Without Losing Context
Writing the third or fourth CRUD endpoint in a day is not a high value cognitive task. It is, however, prone to human error due to boredom or fatigue. Delegating this to an agent allows you to stay in the flow state for the harder problems.
The key advantage of this restricted strategy includes:
- Faster project initialization.
- Consistency in configuration files like Dockerfiles or CI/CD pipelines.
- Reduced friction when switching between different programming languages or frameworks.
- Easier code reviews since the generated code follows standard, predictable patterns.
Where Agents Excel in Scaffolding
Specific areas of development are particularly suited for this approach. These are tasks where the requirements are clear and the implementation is largely standard across the industry.
Infrastructure and Configuration
Setting up a new project often involves creating package.json files, .gitignore entries, environment variable templates, and containerization scripts. These files are critical but repetitive. An agent can generate a functional Docker Compose file or a Kubernetes manifest in seconds based on your project description.
Unit Test Shells
While you should write the logic of your tests, an agent can generate the shells for those tests. It can mock the necessary dependencies and create the test suites, leaving you to fill in the specific assertions and edge cases.
Data Transfer Objects and Schema Definitions
Defining types for your API responses or database schemas is a prime candidate for automation. If you provide a JSON sample, an agent can quickly generate TypeScript interfaces or Pydantic models. This ensures your code is type safe without the manual effort of mapping every field.
Integrating Agents into Your Local Workspace
To make this strategy effective, the tool you use must support a fast feedback loop. Many developers find that switching between a browser and a terminal breaks their concentration. This is where specialized environments like Deska become useful.
In a dedicated workspace, you can run multiple panels side by side. For example, you can have a terminal open for your build process while a coding agent panel generates the boilerplate you need. This layout allows you to see the generated code and the execution results simultaneously.
| Task Category | Agent Role | Developer Role |
|---|---|---|
| Project Setup | Generates folder structure and config | Reviews and refines architecture |
| API Development | Creates boilerplate routes and controllers | Implements business logic and security |
| Testing | Generates mock data and test suites | Defines test cases and assertions |
| Documentation | Generates docstrings and README drafts | Verifies accuracy and adds nuance |
The Role of Local First Tools
Privacy and security are often the primary concerns when integrating AI into a professional workflow. Using a local-first approach ensures that your proprietary code and file structures remain on your machine. Tools that allow you to bring your own API keys or use managed inference give you flexibility in how you consume AI services.
When using tools like Claude Code or Codex CLI within Deska, the agents operate directly on your local files. This proximity reduces the friction of moving code from a chat window into your project. You can ask the assistant to generate a boilerplate panel, verify it in the built-in browser, and then refine it manually in the editor.
Managing the Output of Scaffolding Agents
Even when using agents only for boilerplate, you must verify the output. The goal is to speed up the process, not to skip the validation step.
- Review the Imports: Agents sometimes suggest outdated or unnecessary libraries. Ensure the generated code aligns with your project standards.
- Check for Security Defaults: Boilerplate often uses default settings that might not be secure enough for production environments.
- Maintain Naming Conventions: Ensure the agent follows your specific casing and naming patterns to keep the codebase cohesive.
For those who need to step away from their main machine, a mobile companion can be useful. It allows you to monitor long running agent tasks or review generated code through a secure relay without exposing ports. This is particularly helpful when an agent is generating a large volume of scaffolding that you want to check before finishing your work session.
FAQ: Using Agents for Boilerplate
Is it safe to use agents for boilerplate in commercial projects?
Yes, provided you use tools that respect your privacy and data ownership. Using a local-first environment ensures your source code is not used to train global models without your consent. Always review the generated code for security vulnerabilities.
Which agents are best for generating project structures?
Several options exist, such as Claude Code and OpenCode. These can be run side by side in a unified workspace to compare their outputs. The best agent is often the one that has the most up to date knowledge of the framework you are using.
How do I prevent agents from overwriting my manual changes?
It is best to use a modular approach. Direct the agent to generate new files or specific sections of code rather than giving it full control over existing logic. Using Ask Deska to drive the workspace can help you open specific panels and run commands to isolate the agent activity.
Start Building with Deska
If you are looking for a way to integrate multiple AI agents into your daily routine without losing the benefits of a local environment, Deska offers a flexible solution. The infinite canvas allows you to organize your terminals, editors, and agents in a way that fits your specific workflow.
You can download the free desktop app for Mac, Windows, or Linux and start organizing your workspace today. The workspace itself is free, and you can choose between using your own API keys or managed inference.
Visit our download page to get started with a more efficient way to manage your development projects.