review-llm-artifacts

To detect technical debt from LLM coding agents, you can scan your codebase to identify and categorize specific artifacts like over-abstraction, dead code, and verbose testing. This process uses parallel subagents to analyze tests, dead code, abstractions, and style.

During a code review for LLM-generated code artifacts, look for DRY violations, unnecessary mocks, defensive coding overkill, dead code, over-engineering, and stylistic verbosity. These common anti-patterns reduce maintainability and require targeted refactoring.

To scan your current directory for LLM coding artifacts, run a specialized review skill that analyzes your codebase and generates a structured JSON report. This report includes a summary table to help prioritize refactoring efforts based on risk and safety.

Yes, you need git installed to determine the scope of changes for analyzing LLM-induced issues, and jq is required as a dependency. Git helps establish the review boundaries while the tool processes the structured JSON output for automated refactoring workflows.

Yes, you can automate refactoring workflows to clean up LLM-generated code by using the structured JSON output provided by the analysis. This output categorizes testing flaws, dead code, over-engineering, and stylistic issues for automated remediation.

Static analysis for LLM artifacts operates across multiple programming languages by spawning specialized subagents. These subagents perform parallel, focused reviews on abstractions, dead code, tests, and style regardless of the specific language framework.