Code Generation Repair

Overview

This agent is designed to automate the tedious process of debugging and repairing code generated by other tools. By analyzing a dedicated failure report file (e.g., /output/tests/failures.json), it pinpoints exactly which parts of the generated package are failing tests.

Its core function is surgical: it reads the failures, understands the context of the existing valid code, and applies the minimal necessary changes to patch only the broken sections of the codebase.

Capabilities

• Failure Analysis: Reads structured test failure reports to identify precise points of failure.
• Context Preservation: Ensures that all previously validated and working sections of the generated package remain untouched.
• Minimal Patching: Applies the smallest possible code modifications required to satisfy the failing unit tests, preventing over-rewriting.
• In-Place Editing: Rewrites only the affected files within the generated code structure directly.
• Structured Logging: Emits detailed progress updates to standard output using JSON format for easy integration into CI/CD pipelines.

Example Use Cases

1. Automated Build Correction: After a large code generation run fails due to integration bugs, feed the failure report to this agent to automatically generate a patch file that fixes all reported test failures.
2. Iterative Development Loop: Integrate this into a continuous development cycle where initial scaffolding is generated, tested, and then refined by running this repair agent until all tests pass.
3. Debugging Complex Systems: Use it when manual debugging of large, interconnected code modules becomes overwhelming; the agent narrows the focus to only the failing components.