screenshot-interaction-analyzer — Claude Code subagent: UI flows
A subagent that reads a UI screenshot and maps what can be clicked, where navigation leads and which states change. Useful when you have a screen image but no spec.
npx claude-code-templates@latest --agent ui-analysis/screenshot-interaction-analyzerWhat screenshot-interaction-analyzer does
The subagent acts as an interaction designer. It looks at a screenshot and splits it into layers: clickable elements, input fields, navigation, state transitions and feedback patterns such as loading indicators or confirmation dialogs. Primary actions get a priority label — high, medium or low.
The result is structured JSON. It contains primary_actions, a navigation block with the user's current location, input_flows with fields and submission method, state_transitions as trigger-and-result pairs, and user_journeys listing possible paths. That shape is easy to hand off to tests, layout tickets or a prototype description.
The toolset is narrow: Read and TodoWrite. So it reads the image and keeps a task list, but never opens project code — checking the analysis against real markup is a separate step. The agent file lands at .claude/agents/screenshot-interaction-analyzer.md.
What it helps with
- break a screenshot into clickable elements and rank them
- reconstruct navigation and locate the current screen in it
- list form fields and how the form is submitted
- write down state transitions triggered by a click or tap
- sketch possible user journeys from a single screen
Example prompts for the agent
- analyze this checkout screenshot and return JSON with actions and transitions
- look at this screenshot and list the input fields and filters on it
- map the navigation in this screenshot and list possible user paths
How to install screenshot-interaction-analyzer
1claude-code-templates
npx claude-code-templates@latest --agent ui-analysis/screenshot-interaction-analyzer2Manually: put the files in place
.claude/agents/screenshot-interaction-analyzer.mdFor all projects use ~/.claude/agents/. Claude delegates by the description field or an @-mention
Manually: put the files in place
.codex/agents/screenshot-interaction-analyzer.tomlCodex defines subagents in TOML: name, description, developer_instructions. Move the Markdown text into developer_instructions by hand. For all projects use ~/.codex/agents/
Manually: put the files in place
.cursor/agents/screenshot-interaction-analyzer.mdCursor also reads .claude/agents/
Step by step, with commands for each platform: Claude Code subagents: how to create one, call it and keep context clean
Questions about screenshot-interaction-analyzer
How do I install screenshot-interaction-analyzer in Claude?
Use the templates CLI: run npx claude-code-templates@latest --agent ui-analysis/screenshot-interaction-analyzer, then the file .claude/agents/screenshot-interaction-analyzer.md is added to your project.
What does screenshot-interaction-analyzer do?
It analyzes a UI screenshot and returns JSON covering clickable elements, navigation, input fields, state transitions and possible user journeys.
Which tools does the subagent have?
Only Read and TodoWrite. It reads the image and tracks tasks, but does not open or edit project code.