fix-llm-artifacts

What This Skill Does

The fix-llm-artifacts skill provides an automated, structured approach to cleaning up technical debt and refactoring artifacts left behind by LLM-assisted coding sessions. It acts as a safety-conscious mediator that interprets the output of prior review cycles (via review-llm-artifacts), categorizing findings into 'Safe' and 'Risky' buckets. Safe fixes, such as removing unused imports or clearing stale comments, can be applied automatically, while risky changes involving behavioral logic or complex abstractions require human verification. This ensures that your codebase stays clean without risking inadvertent regression.

Installation

You can install the skill directly through the ClawHub CLI using the following command: clawhub install openclaw/skills/skills/anderskev/fix-llm-artifacts Ensure your current working directory is the root of your project, as the tool relies on Git state validation to prevent applying patches to stale code versions.

Use Cases

• Post-Refactoring Cleanup: After a series of rapid AI-assisted iterations, use this skill to mass-remove redundant type annotations and overly verbose docstrings.
• Test Suite Optimization: Systematically address test_refactor suggestions to consolidate test boundaries and remove dead code paths that were bypassed during iterative prototyping.
• Codebase Auditing: Run with the --dry-run flag to generate a comprehensive report of potential improvements across your entire project architecture.

Example Prompts

1. "/beagle-core:fix-llm-artifacts --category tests --dry-run"
2. "/beagle-core:fix-llm-artifacts --all"
3. "/beagle-core:fix-llm-artifacts --category style"

Tips & Limitations

• Git Safety: Always ensure you have a clean working tree before running the tool. While the agent automatically stashes uncommitted changes, performing a manual commit or push prior to execution is a best practice.
• Stale Results: If you have modified your code since the last review, the tool will notify you. Do not force apply fixes on stale results, as line numbers and context blocks will likely have shifted, leading to failed patches.
• Risky Fixes: The tool categorizes structural changes as 'Risky'. We highly recommend running these in small batches and using your standard test runner immediately afterward to confirm the integrity of the logic changes.