BUILD YOUR OWN AUTOFIX WITH PATCHWORK

Summary from the TLDR InfoSec newsletter

This post describes a Patchflow to create an AI code autofixer. The Patchflow first scans the code using Semgrep. It then enriches the Sarif output with the source code file and uses an LLM to triage the report. The LLM is then prompted to create and test a patch. If successful, the LLM then opens a PR with the fix. Prompts for each of the LLM steps are included.