Using the Tone of Voice Skill with Your Agent

What is this?

The Tone of Voice skill gives your agent a clear instruction on how a text should sound. You can use it to have a finished answer checked and adjusted against your tone as a final step. Unlike before, a skill like this no longer runs automatically the way Shared Instructions did, always working in the background. You have to define when it should become active.

How the skill works

Think of the Tone of Voice skill less as extra information and more as a command. The agent runs it as its own step instead of just reading it as background knowledge. That's why it doesn't necessarily trigger just because it exists. It kicks in when you call it, or when the agent decides during the run that it needs it.

Activating the Tone of Voice skill manually

You activate the skill directly with the slash command.

1. At the point where the skill should apply, type /.

2. Select General Tone of Voice.

3. The agent runs the instruction and uses it to check the final answer against your tone.

⚠️ Setting the skill with / doesn't automatically mean the result gets better. That depends on how the skill is written. A skill that's just called "General Tone of Voice" gives the agent no clear command. Write it as an instruction to act, for example: "Check the existing text for tonality XY and adjust it accordingly." That's what leads to a real improvement.

Letting the agent pull the skill dynamically

You don't have to set the Tone of Voice skill with a slash. If you write "let's check this with the Tone of Voice" in your instruction, the agent pulls the skill dynamically when it thinks it needs it. That works too.

Seeing what happened

You can always check whether and when the skill ran.

• The history shows that the "Tone of Voice" skill was activated, for example to proofread the final answer.

• The agent's context or thought-process window also shows when it pulled the skill dynamically.

When the Tone of Voice step makes sense

An extra step at the end, like "check the result with the skill", is worth it when the agent has a specific, recurring tonality problem and the skill is written as a clear instruction to act.

When NOT to use it

If the agent already hits the right tone and the answers read well, you don't need the extra step. More context doesn't automatically make the result better. Quite the opposite. Too much of it can make things worse instead of better.

Good to know: Where an extra Tone of Voice step actually makes sense and where it doesn't is something we're evaluating together with customers right now. If you're unsure whether it helps your setup, reach out. We'll take a look together.

You're all set and good to go!