Disclosed weakness · no CVE recorded

An AgentScope text-embedding cache identifier could include a provider key

AgentScope · AgentScope · DashScope embeddings

First documented disclosure · Review checkpoint

Summary

The DashScope text embedding adapter included a per-call API key when it constructed an embedding cache identifier.

Severity and exploitation evidence

No CVSS score recorded in this review.

Exploitation not established: The public report and upstream fix document the defect. They identify no malicious use.

Exploitation evidence source

Technical severity is not a measure of observed attacks or of risk to your deployment. Attempts, demonstrations, and successful compromise retain the qualifications in the source.

How the weakness works

The text adapter copied request options into a cache identifier without first excluding the secret.

Conditions required

An application supplies a per-call DashScope key on the text path and uses a caching backend.

Potential impact

A cache backend that retains the identifier could expose the credential to people or services with cache access.

Affected versions and configurations

The project verified the pre-fix implementation. A complete affected release range was not published.

Fix and mitigation status

2.0.12 includes the correction.

Practical next steps

Upgrade to 2.0.12 or a later fixed version. Review retained identifiers and rotate keys found in accessible cache records.

Operating lesson

Exclude provider credentials before passing request metadata to a cache.

Evidence and limits

The October 24 issue and patch document the identifier-level exposure. External access to a cache was not observed. The November 16 release lists the patch.

Sources