Codemod Runner Agent

What Codemod Runner Agent does

A codemod runner agent is an AI agent specialized in codemod runner. It does not answer from memory alone. It read files, apply patches, and run tests, stores observations, verifies completeness, and delivers verified code change set with explicit gaps when tools fail.

The agent receives a codemod runner task, plans tool calls (repository, terminal, test-runner, linter), executes read files, apply patches, and run tests, keeps working notes, verifies the draft, and publishes verified code change set. If test suite failed after the patch was applied, it revert the failing hunk and iterate with a smaller change. This page simulates that observable loop — not private chain-of-thought.

Workflow

Each step maps to a node in the lab above. Scroll up to play and watch the packets travel.

1. 01

   Input

   Receive task

   Accept the codemod runner request and output constraints.

   Data preview

   In

   Run codemod runner for Q3: produce verified code change set with audit trail.

   Out

   Normalized objective: produce verified code change set.

2. 02

   Plan

   Plan the run

   Choose tool order for Read files, apply patches, and run tests.

   Data preview