clqms-be/.agent/artifacts/how_to_control_agent_behavior.md

# How to Control Agent Behavior

## Stopping Auto-Verification After Implementation

If you want the agent to **stop verifying/testing** after implementing code in future chats, you can:

### Option 1: Add to User Rules (Recommended)

Go to your **User Settings** and add a custom rule:

```
After implementing code changes, do not automatically verify or test unless explicitly asked.
```

This will apply to all future conversations.

### Option 2: Use Explicit Instructions

In each chat, you can say:

- ✅ "Implement X **without testing**"
- ✅ "Just create the code, **don't verify**"
- ✅ "Skip verification, I'll test it myself"
- ✅ "**No auto-run** after implementation"

### Option 3: Workflow Annotation

Add `// turbo` or `// turbo-all` annotations to workflows you want auto-executed.

For workflows you **don't** want auto-executed, simply don't add the annotation.

---

## Current Behavior

By default, the agent is **proactive** and will:
- ✅ Implement your request
- ✅ Run commands if safe
- ✅ Verify the implementation works
- ✅ Test and report results

This is helpful for catching errors early, but you can disable it with the methods above.

---

## Example User Rule

Add this to your **User Settings → Custom Rules**:

```
Implementation Preference:
- After creating or modifying code, wait for my explicit instruction before testing
- Do not auto-run commands unless I specifically ask
- Provide a summary of changes and ask if I want to test
```

---

## Quick Reference

| What You Want | How to Achieve It |
|---------------|-------------------|
| No auto-testing | Add user rule: "Don't auto-test" |
| Sometimes test | Use explicit instructions per request |
| Always test specific workflows | Add `// turbo-all` to workflow |
| Never test specific workflows | Don't add turbo annotations |

---

**Note:** The agent will still be helpful and provide guidance, but won't automatically run verification commands unless you ask.