Getting Started with SiloBrief: Secure Context Export for AI Analysis

Start by cloning the repository and installing dependencies. Since SiloBrief operates in high-security contexts, ensure your environment is isolated. Run `git clone https://github.com/d3vksy/silobrief.git` and navigate to the directory. Use `pip install -r requirements.txt` to set up the Python environment. Initialize your project by running `silobrief init --profile airgap`. This command creates a local configuration file `silobrief.yaml` where you will define security policies and export rules. Verify the installation by running `silobrief --version` to ensure the CLI is accessible in your PATH.

Open `silobrief.yaml` to configure what data can be exported. This is the core security mechanism. Define `include` patterns for specific file paths or function signatures. For example, add `include: ['src/core/auth.py', 'api/handlers.py']` to restrict exports to critical modules. Set `exclude` for sensitive data like secrets or database credentials using patterns like `exclude: ['*.env', 'config/secrets.json']`. You can also specify `redact` rules to mask specific strings, such as API keys, using regex patterns. Save the file and validate the syntax with `silobrief validate`. This step ensures that only necessary context is exposed to the less restrictive environment.

Run the extraction command to generate the secure brief. Use `silobrief export --target ai-analysis` to create a sanitized package. The tool scans the codebase, applies your YAML policies, and generates a `brief.json` file containing only the allowed code snippets, function signatures, and architectural metadata. This file is structured for AI consumption, including context headers that explain the relationship between extracted components. Check the output directory for the generated file and any associated logs. If the export fails, review the error messages for policy violations or syntax issues in your code. The process is non-destructive and does not modify the original source code.

Move the `brief.json` file to your AI-assisted development environment. If using a local LLM setup, load the JSON into your prompt context. For cloud-based AI tools, upload the file securely. The structured format allows AI models to understand the code architecture without accessing the full, sensitive repository. You can now ask the AI to analyze specific functions for bugs or suggest optimizations based on the provided snippets. Ensure that the AI tool you use respects the same security boundaries and does not retain the data longer than necessary. This step bridges the gap between secure development and external analysis capabilities.

Analyze the AI's feedback and update your export policies if necessary. If the AI lacks sufficient context, adjust your `include` patterns to add more relevant files or function signatures. Conversely, if sensitive data was inadvertently included, refine your `exclude` or `redact` rules. Re-run the export and repeat the analysis process. Over time, you will develop a robust set of policies that balance security with the need for external insights. Document your successful policy configurations for future projects. This iterative approach ensures that your security posture remains strong while leveraging AI for efficient debugging and code improvement.