Airgentic Help
Reusable Prompt Components keep identity, tone, safety, and formatting consistent across every agent. Agent prompts include them with substitution tags such as {{prompt:Tone}}.
The Edit Prompts screen lists components in their own section. Inherited (global) components apply to all services until you edit one, which creates a service-specific copy.
| Component | Typical contents | Include with |
|---|---|---|
| Identity | Organisation name, agent persona, what the service does | {{prompt:Identity}} |
| Tone | Communication style (formal, friendly, concise) | {{prompt:Tone}} |
| Performance Style | How the agent refuses poems, roleplay, fake accents, and other performative formats | {{prompt:Performance Style}} |
| Guardrails | Scope, safety, and topics to decline or redirect | {{prompt:Guardrails}} |
| Output Format | Response structure, citations, length | {{prompt:Output Format}} |
New agent prompts created from the standard template already include Identity, Tone, Guardrails, Output Format, and Information Retrieval. Tone already includes Performance Style, so you do not need to add that tag to each agent prompt.
Performance Style is the shared rule that the agent must answer in professional customer-service language. It covers user requests for poems, rhymes, songs, jokes, stories, roleplay, character voices, theatrical styles, and similar creative formats.
If the user also asked an in-scope factual question, the agent should still answer that question — in normal prose, not in the requested performance format.
There are two Performance Style files:
| File (on Edit Prompts) | Used when |
|---|---|
| Performance Style | Text chat |
| Performance Style Voice | Voice mode (including telephony) |
The tag is always {{prompt:Performance Style}}. During a voice session the platform substitutes the voice variant automatically. Do not insert {{prompt:Performance Style Voice}} into agent prompts.
The voice variant tells the model to keep the speaking accent, pace, and voice character already set for the session (in the voice preamble). Rules against “accents” and “character voices” apply to user-requested performances, not to the configured voice.
Put {{prompt:Performance Style}} in Tone only. The global Tone component already has it.
Do not also add it to Guardrails or to individual agent prompts. Duplicating the tag (or copying the same anti-performance wording into Guardrails) can fight the voice variant — for example, a second copy that says “never use an accent” can flatten a configured regional speaking voice.
| What you want to change | What to edit |
|---|---|
| Poem / roleplay refusal in text chat | Performance Style (creates a service-specific copy) |
| Spoken accent, pace, or “keep this voice” rules | Performance Style Voice, and the voice preamble |
| Formal vs friendly wording | Tone — leave the Performance Style tag in place |
Example: an Australian council that wants a clearer Australian speaking voice can override Performance Style Voice for that service, and adjust the voice preamble. Text chat still uses the global Performance Style unless you override that too.
Leave the {{prompt:Performance Style}} line in Tone when you customise Tone. If you delete it, text and voice lose the shared performance rules (unless you paste equivalent wording yourself).