The boundary of a smart contract
A smart contract can read information made available within its execution environment. External events, such as an exchange price or weather observation, need a mechanism to enter that environment. An oracle provides that connection.
An oracle can supply data, deliver messages or support other forms of off-chain interaction. It is part of the application’s trust model rather than a magical guarantee that an external claim is true.
Price feeds need context
Check where observations originate, how they are aggregated, how often updates occur and what happens when a source fails. A price can be accurate at one instant but unsuitable after a long delay.
A lending protocol may use an oracle price for liquidations. Manipulating a thin market that feeds a poorly designed oracle can therefore affect more than the traded asset itself.
Ask the failure questions
Imagine a feed stops updating during a volatile period. Does the application reject stale data, pause selected operations or continue using the previous observation? Each choice has consequences.
Also inspect who can replace a feed or change its parameters. An application’s security depends on the relationship between its contracts and its data inputs. When reading incident coverage, distinguish a compromised data source from a bug in how an application used otherwise valid data.
Why the observation time matters
Suppose a lending contract values collateral using an oracle observation of $100, while accessible markets have moved to $85. The gap can affect borrowing capacity and liquidation decisions. This hypothetical situation illustrates stale-data risk; it does not establish that any particular oracle would accept that price.
An oracle consumer should account for the feed’s intended use, decimals, update conditions and timestamps. A number returned successfully by software can still be unsuitable for the decision being made. Different feeds may represent different pairs or calculation methods.
Separate the data source from its delivery
A design can aggregate observations from multiple venues and use multiple participants to publish a result. Diversity at one layer does not prove independence at every layer: several publishers might ultimately rely on overlapping underlying markets.
Read what happens when updates stop, observations diverge or the feed is paused. For a reported oracle incident, identify whether the problem involved market manipulation, stale input, delivery or the consuming contract’s assumptions. “The oracle failed” is too broad to explain the cause by itself.
Sources & further reading
Educational content. Examples are illustrative. Protocols, products and rules can change; consult the linked documentation for current details.