Impact
A flaw in the internal/memory/doc.go discoverDocs function allows input that is not properly neutralized to be sent to the language model prompting engine. If an attacker supplies specially crafted data, the LLM may generate code or instructions that can be executed by the host, or otherwise manipulate the behavior of the system. The vulnerability could lead to unauthorized data exposure, tampering, or execution of malicious commands; it also undermines the integrity of generated responses.
Affected Systems
esengine DeepSeek‑Reasonix up to and including version 2.52.0 is impacted. The fix is available in release 2.53.0, identified by commit 4ca33493a673c2788aa8598fbca2b2f47a69e1b8.
Risk and Exploitability
The CVSS score of 5.3 indicates moderate severity. The EPSS score is not available, and the vulnerability is not listed in the CISA KEV catalog. Attackers can trigger the flaw remotely, as the component exposes a network interface that accepts prompting input. Publicly disclosed evidence shows that exploits can be constructed, so timely patching is recommended.
OpenCVE Enrichment