Loop

Review and rework loop

You might say: evaluator loop, fix and recheck, ping-pong

An automated review fails a pull request, a rework run fixes what the review named and pushes, and the push starts a new review. A try cap ends the loop and hands the pull request to a person.

What it does

Fixes review findings without a human, and stops a disagreement between two runs from looping forever.

Ask

How many tries does the loop get before a human, and what does the human see when it stops?

Trouble

The same finding comes back on every try, or the loop runs all night on one pull request.

Loop: balancing

Automated review → Run → Automated checks → Try cap ↺

Pattern

Accelerator with a brake

Connections

• See also: Try cap

Sources

1. Anthropic, Building effective AI agents, Workflow: Evaluator-optimizerIn the evaluator-optimizer workflow, one LLM call generates a response while another provides evaluation and feedback in a loop.
2. Anthropic, Building effective AI agents, AgentsAgents can then pause for human feedback at checkpoints or when encountering blockers.
3. Anthropic, Building effective AI agents, Agentsit’s also common to include stopping conditions (such as a maximum number of iterations) to maintain control.

Use it on your project

teardown-apply: hold "Review and rework loop" (review-rework-loop) from the teardown at https://teardownatlas.com/software-factory/ against my project

Paste it into a coding agent that has the teardown skills.

An AI agent wrote this map with the teardown skill. It can contain errors: check the sources before you rely on a part. About and legal