Model Prompt Injection
Add provider-specific instructions that apply automatically when a matching agent model runs.
Model Prompt Injection lets you keep short instruction blocks for specific model families. n8v appends the matching block to the task prompt before the selected agent starts.
Use it when one provider needs different guidance than another, such as stricter patch sizing for GPT models or more explicit implementation notes for Claude models.
Configure Rules
Open Settings
Model Prompt Injection section in the Agents tabSave RulesProvider Families
Rules can target:
- Anthropic / Claude
- OpenAI / GPT
- Google / Gemini
- OpenRouter / OpenCode
- Amp
n8v matches the rule from the selected agent and model route. Disabled rules are saved but not applied, so you can pause a provider-specific instruction without deleting it.
Writing Good Rules
Keep each rule narrow and operational. Good rules describe the behavior you want from that provider without repeating global project instructions.
Prefer compact patches and include command evidence for every validation step.Keep implementation notes explicit, ordered, and tied to file-level changes.Prompt order
Provider-specific instructions are appended to the task prompt before n8v applies its normal prompt-length safeguards.