# Write Personality

Source: https://docs.generalaugment.com/guides/write-personality/
Description: Configure SOUL.md and behavior settings.

Agent personality lives in `SOUL.md` and maps to the project's system prompt.

```markdown
---
name: Demo Agent
role: Product support assistant
tone: concise, proactive, friendly
---

# Demo Agent

You help users answer product questions and prepare clear next steps.

## Rules

- Confirm before taking any account-changing action.
- Keep links and next steps specific to the user's product area.
- Keep chat replies under 200 words unless detail is requested.
```

Deploy updates:

```bash
genaug dev ./demo-agent/genaug-agent.yaml --message "Help me prepare onboarding notes"
genaug deploy ./demo-agent/genaug-agent.yaml
genaug onboarding verify --project demo-agent --json
```

Before launch, prove the hosted runtime actually changed behavior. A passing deploy
only proves the files were accepted; smoke or trace evidence should show the next
`/v1/responses` turn following the updated SOUL rules. Trace and support-bundle
artifacts now include skill names loaded or created by Hermes when that evidence is
available; if final behavior changes but no trace/support-bundle evidence links it to
SOUL or skills, keep that as a launch-proof gap.

Python test:

```python

response = httpx.post(
    "https://api.generalaugment.com/v1/responses",
    headers={"Authorization": "Bearer gaadmlive_your_project_scoped_key"},
    json={"model": "balanced", "input": "What is your style?"},
    timeout=30,
)
response.raise_for_status()
data = response.json()
print(
    next(
        part["text"]
        for item in data["output"]
        if item["type"] == "message"
        for part in item.get("content", [])
        if part.get("type") == "output_text"
    )
)
```

See [Quickstart](/getting-started/quickstart/) for the full deploy loop.
