Separate what is observed from what is inferred.
Every 8thGuard finding names its source, publishes its rule version, and keeps missing coverage explicit. No black-box scores. No hidden assumptions.
Four layers, each with a stated limit
- 01
Input integrity
Validate the submitted public identifier and its chain. Limit: a valid-looking format proves neither existence, ownership nor safety.
- 02
Public-chain evidence
Status, activity, asset movement, counterparties, permissions. Limit: public-chain facts show what happened, not why, nor who controlled an address.
- 03
Supplied context
A caller may supply additional context. It is not independently established merely because it appears in the request. Limit: supplied context can be incomplete, manipulated or impossible to verify.
- 04
External evidence
Configured sanctions feeds, public blocklists, curated evidence and supported vendor observations. Coverage identifies unavailable or incomplete sources. Limit: absence from a list is not proof of safety.
Unknown is a valid result.
Available
The source responded and the observation can be described.
Partial
Some evidence exists, but important dimensions remain uncovered.
Unavailable
The source could not be checked, so no claim is made from it.
How to read an outcome
- Obs
Observation
A directly visible fact from a named source or submitted material.
- Ind
Indicator
Why one or more observations may matter for this decision, with a severity and the rule that fired. Never a numeric score, never a label on a person.
- Unk
Uncertainty
What cannot be established from the available data and should remain unresolved.
- Next
Next step
A proportionate check, pause, clarification or decision the reader can consider.
Rules are published, never edited.
Each answer records its rules version and issue time. Changing a rule creates a new version. Source data can change too: the same identifier and rules version do not guarantee an identical result later. Keep the issued response when you need to understand a past decision.