Manage LLM prompts and skills
OctoMY sends prompts to language models — for example, Trove's implement and verify prompts that drive its autonomous coding sessions. You can read, edit, and reset every one of these prompts, and you can add optional skills that extend a session's standing instructions.
Pick a task:
| I want to… | Guide |
|---|---|
| Change what a prompt says, then undo my change | Edit a prompt |
| Turn on an optional skill for a Trove project | Enable a skill |
How prompts resolve
Each prompt has a stable key (for example core.trove.verify). Its active text comes
from a three-tier cascade, highest priority first:
- Project — an edit made for one Trove project.
- App — an edit made for this app, shared by every project.
- Factory — the version shipped inside the app. Always present. Read-only, but you can override it at the App or Project tier.
The manager shows a badge on each prompt (Factory, App, or Project) so you can see where its active text comes from. Reset to factory removes your override and reveals the tier below.