Credential sent in a request, not as authentication
A credential seen in one call is sent by a later call in a request body, URL or argument to an unrelated host.
Why it is risky
Section titled “Why it is risky”A token that appears in one call and is then sent in the body or URL of a request to an unrelated host has left the place it was meant to be used.
Example
Section titled “Example”An agent sets a GitHub token in its environment for a build, then posts that token as form data to an unfamiliar web service.
This is a fictional example, not an event from a customer.
What Omnodex Detects
Section titled “What Omnodex Detects”This advanced rule notes a recognized credential value in one call’s parameters and fires when a later call in the session sends the same value in a request body, URL or argument to a host that is not the credential’s own service. Sending a token as an authentication header, or to its own service, is not flagged.
| Rule ID | Event category | Tier | Base severity |
|---|---|---|---|
RULE_ADV_CREDENTIAL_VALUE_SENT | credential_movement | Requires Pro | CRITICAL |
Base severity is the rule default. A finding involving content saved for later execution may have a lower severity. Your loaded rule version and event details determine the finding you see.
What to check next
Section titled “What to check next”Check the destination and the request. Treat the credential as exposed if the request reached a host you do not control, and rotate it.
False positives
Section titled “False positives”Some legitimate tools expect a token as a request parameter rather than a header, and a service can be reached under a host name that is not recognized as the credential’s provider.
False negatives
Section titled “False negatives”Only recognized credential formats are tracked. A value that was transformed before sending, split across calls, or sent in another session is missed.
Technical details
Section titled “Technical details”The value is held in memory only while the session is analyzed. It is never written into the finding, which names the credential by kind and how it was sent. This advanced rule produces evidence with the finding: each call in the chain with its role, a one-line summary of what it did, where the last call went, and why the chain might be routine. In the hosted dashboard, related findings about the same chain are read together as one investigation. Availability depends on the advanced rule bundle loaded by the installation. Findings on activity captured by agent hooks appear when the session ends or at the next background pass, within about 15 minutes; activity through the MCP proxy is judged as it is recorded.
A finding describes a pattern in captured activity. Review the original calls, their results, and the task context before concluding that harm occurred. Absence of a finding does not establish that an operation is safe.