AI-native Development Workflow

AI-native development with an AI agent proceeds from defining specifications through packaging as a user module, repeating generation and verification along the way.

Basic Flow

1. Start the AI agent
2. Define the specifications for the artifacts you want to generate
3. Create a prompt based on the following criteria and instruct the AI agent
   • Clearly state the specifications (if in a separate file, specify the file location)
   • Instruct the AI agent to generate artifacts according to the skill sets
4. The artifacts are saved under src/
5. Deploy the artifacts to the intra-mart verification environment and confirm they work
6. If there are issues, ask the AI agent to fix them
7. Repeat steps 4–6 until the artifacts work correctly
8. Once the artifacts are complete, package them as a user module and register it in the IM-Juggling project

AI Operation Permissions During Artifact Generation

While work is being delegated to the AI agent, the agent may request permission for the following operations. When prompted, confirm that the operations are within the project directory before granting permission.

• Saving files
• Executing Node.js scripts

Improving Accuracy

When generation accuracy does not improve as expected, the following approaches can help.

If the prompt or specification document is too large, the context may overflow, causing some specifications or instructions to be ignored and reducing artifact accuracy. If accuracy drops, asking the AI agent to re-verify consistency with the specifications and rules after generation can help correct the artifacts.

Please review the generated artifacts for any inconsistencies with the provided specification document and fix them if found.
Also, check the artifacts according to the checklist/validation script.

When requesting bug fixes, include any exception messages in the prompt. Whenever possible, include error messages directly related to the cause, as well as the file path and line number where the error occurred, to improve fix accuracy.

When submitting a request from the application screen, the following error occurred. Please investigate the cause and fix it.
```
log.message=Failed to retrieve the last process task ID. System case ID: ma_8hz8171jnvj0aqh, Action node ID: lv_req_apply

jp.co.intra_mart.system.workflow.engine.common.EngineException
  at jp.co.intra_mart.system.workflow.engine.tool.EngineMatterUtil.getLastProcessID4ActiveMatter(EngineMatterUtil.java:34)
  at jp.co.intra_mart.system.workflow.engine.action.AbstractActionEventBase.removeFlowXMLCache(AbstractActionEventBase.java:2530)
  at jp.co.intra_mart.system.workflow.engine.action.AbstractActionEventBase.fire(AbstractActionEventBase.java:336)
  at jp.co.intra_mart.system.workflow.internal.manager.action.ActionManager.apply(ActionManager.java:127)
  at jp.co.intra_mart.foundation.workflow.application.process.ApplyManager.apply(ApplyManager.java:421)
  at jp.co.intra_mart.system.workflow.javascript.application.process.ApplyManagerObject.jsFunction_apply(ApplyManagerObject.java:282)
  at java.base/jdk.internal.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
  at java.base/jdk.internal.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:62)
  at java.base/jdk.internal.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43)
  at java.base/java.lang.reflect.Method.invoke(Method.java:566)
  at jp.co.intra_mart.system.javascript.MemberBox.invoke(MemberBox.java:126)
  at jp.co.intra_mart.system.javascript.FunctionObject.call(FunctionObject.java:442)
  at jp.co.intra_mart.system.javascript.optimizer.OptRuntime.call2(OptRuntime.java:42)
  at _im_95_workflow._common._proc._exec._apply_95_jssp_46_js._c_apply_1(%USERPROFILE%\resin\webapps\imart\WEB-INF\jssp\platform\src\im_workflow\common\proc\exec\apply_jssp.js:163)
  at _im_95_workflow._common._proc._exec._apply_95_jssp_46_js.call(%USERPROFILE%\resin\webapps\imart\WEB-INF\jssp\platform\src\im_workflow\common\proc\exec\apply_jssp.js)
```

When Using MCP

If you are connected to the iAP development environment using the AI-native Development Support Feature and MCP, the exception logs are retrieved via MCP, so you do not need to paste them into your prompt. For the installation steps and usage of the AI-native Development Support Feature, see AI-native Development Support Feature.

For screen-related bugs, providing a clear description of what is happening and how you want it fixed will improve the accuracy of the fix.

Regarding the application screen, the following issues were found. Please investigate the cause and fix them.
1. The number of days acquired is not updated even when a start date and end date are selected
2. There are no text boxes for start date and end date (text box + calendar picker is preferred for date input)