Guides to visual feedback and AI workflows
AI coding agents now sit between the person who finds a problem and the person (or agent) who fixes it. The handoff format matters more than it used to. These guides explain how to make it work.
Every guide on this hub includes a worked example, a markdown snippet, a sample prompt, or a step-by-step screenshot walkthrough. The goal is for the page to stand on its own as something someone could read once, close the tab, and apply.
Concepts
Definitional pages that answer one question well.
-
What is agent-readable feedback?
The format coding agents can act on without human translation.
-
Markdown screenshots, explained
How inline image references make screenshots portable.
-
Screen capture vs screen recording
When a still is the right artifact, and when video is.
-
Structured feedback for LLMs
Why structure beats free-form when an LLM is reading.
Workflow
Step-by-step guides for getting feedback into specific tools.
-
How to give feedback to an AI coding agent
The mechanics of a feedback loop that actually closes.
-
Why Loom doesn't work with agents
Agents can't watch videos. The artifact has to read.
-
Getting feedback into Cursor
Composer paste vs. @file reference. What works, what doesn't.
-
Getting feedback into Claude Code
feedback.md in the repo root. The file that survives the session.
Reference
Stop pasting screenshots into agent prompts.
Capture the screen, talk through the problem, hand your agent a markdown document it can act on. About a minute from open tab to shareable URL.
Start capturing