Refine AutoFixes with Aikido AI

Aikido allows to automatically fix different types of findings. The Refine with Aikido AI chat panel lets you iteratively adjust AutoFixes using natural language, without leaving the diff view. This is powered by Aikido Agent, the same intelligence behind Ask Aikido.

Use Cases

AutoFix gets you most of the way there. Refinement is useful when:

• The fix is correct but doesn't match your codebase's naming conventions or patterns
• You want to add explanatory comments
• You want to include unit tests
• You want to introduce specific logic, like error handling or logging calls
• The fix needs to account for context not available to the AI
• You want to tighten scope, for example limiting what gets sanitized or validated

How to Refine a Fix

Step 1. Open the diff modal via the View Fix action and click Refine with Aikido AI in the bottom left.

Step 2. Type your instruction in the text field.

Step 3. The diff view updates automatically when the refined fix is ready.

Each instruction you send is retained as context for subsequent refinements, so the AI builds on prior instructions rather than starting fresh.

Tips for Writing Instructions

Be specific about intent, not just the symptom:

• "Thrown an exception instead of returning an empty object"
• "Rename the sanitized variable from safe_input to safeInput "
• "Add a log when the regex matches"

Persisting the context

If you want Aikido to always take into account your prefences while creating an AutoFix, use Custom Context for AutoFixes.