Orbit doesn’t treat every piece of information equally. Here’s how it decides what to trust — and how you control it.
Why trust levels matter
Your agents produce a lot of output. Some of it is speculative research. Some of it is a finished, verified report. Some of it is a locked decision that should never be contradicted. If Orbit treated all of these the same way, agents would give equal weight to a rough first-pass note and a decision you ratified six months ago. That’s how you get agents contradicting your most important rules. Trust levels solve this. Every piece of content in your Vault carries a trust level — and Orbit’s retrieval system weights higher-trust content more heavily when agents query for context before a run.The three trust levels
Exploratory
“This is a rough finding. Use with caution.” The default trust level. Everything your agent files lands here unless you specify otherwise. Exploratory content is kept in the vault for reference, but agents know to treat it as provisional — it might be wrong, incomplete, or outdated. Use exploratory for: first-pass research, working notes, drafts, intermediate outputs, anything that hasn’t been verified yet.Conclusion
“This is a finished, vetted output. It’s reliable.” A conclusion is something your agent produced that you’re confident in — a completed report, a verified lead list, a finished analysis. Agents give conclusions real weight when querying the vault. Use conclusion for: completed deliverables, verified research, outputs you’ve reviewed and approved.Authoritative
“This is ground truth. Treat it as fact.” Authoritative content is the highest trust level in Orbit. It’s injected directly into every relevant agent’s context at the start of each run — before the agent does anything else. An authoritative decision from 3 months ago outranks a fresh exploratory note automatically. When you file something as authoritative, Orbit flags it for your review before it takes effect. You confirm it, and from that point forward every agent reads it as ground truth. Use authoritative for: locked decisions, confirmed facts, rules that must never be contradicted, pricing, architectural decisions, standing instructions.How trust levels affect agent behavior
| Trust Level | Injected into heartbeat? | Searchable? | Agent weight |
|---|---|---|---|
| Exploratory | No | Yes | Low |
| Conclusion | No | Yes | Medium |
| Authoritative | Yes (automatically) | Yes | Highest |
The trust intercept
When an agent tries to file something as authoritative, Orbit pauses and surfaces it in your Inbox for review before it takes effect. You’ll see:- What the agent is proposing to make authoritative
- Whether it conflicts with anything already in the vault
- Options to: Approve (it becomes authoritative), Downgrade (file as conclusion instead), or Reject
Conflict detection
When a new piece of content contradicts something already in the vault at a higher trust level, Orbit flags it instead of silently overwriting. Both versions stay in the vault — the original and the new one — with their provenance intact (which agent filed each one, when, with what reasoning). You see the conflict in your Inbox and decide which version is canonical. This is how Orbit maintains one source of truth without losing the history of how it got there.Setting trust levels
As an agent: When callingorbit_result, set the trust_level parameter:
exploratory, conclusion, authoritative
From your dashboard: You can promote or demote any vault item’s trust level manually from the item’s detail view. Useful for promoting a conclusion to authoritative after you’ve verified it, or downgrading exploratory content that turned out to be wrong.
Source trust vs. content trust
Orbit tracks two separate dimensions of trust:- Trust level — how confident we are in the content itself (exploratory/conclusion/authoritative)
- Source trust — how much we trust the agent or operator who filed it
- Operator (you) — highest trust
- Agent conclusion — high trust
- Agent exploratory — standard trust
- Agent reflection — lower trust