# CVE-2026-126767: OpsTree REMS prompt injection vulnerability in RCA Chat

Source: https://junglewise.ai/threats/cve-2026-126767-opstree-rems-through-1-0-0-contains-a-prompt-injection  
Publisher: Junglewise Threat Intelligence, https://junglewise.ai/threats  
Published: 2026-11-11. Updated: 2026-11-11  
License: CC BY 4.0 with a link back

CVE-2026-126767 · Severity: medium · CVSS 4.3 · Published 2026-11-11

OpsTree REMS is an AI-powered observability platform that uses an AI assistant to help identify root causes of incidents. Authenticated users can craft malicious chat inputs to trick the AI assistant into disclosing hidden system instructions and internal operational directives, potentially revealing how the platform is configured and opening the door to more sophisticated attacks.

Technologies: [OpsTree REMS](https://junglewise.ai/threats/technologies/opstree-rems). Vendors: [OpsTree](https://junglewise.ai/threats/vendors/opstree).

## Quick answers

**Is CVE-2026-126767 exploited in the wild?** No. Junglewise Threat Intelligence has no report of CVE-2026-126767 being exploited in the wild.

**How severe is CVE-2026-126767?** Medium severity, with a CVSS score of 4.3, published 11 November 2026.

**Which products and versions are affected?** OpsTree REMS through 1.0.0.

**Is there a fix for CVE-2026-126767?** A patch was published: Remediated by OpsTree Solutions; specific version not disclosed Until you apply it: Implement strict separation between system, developer, retrieval, and user-controlled prompts; prevent user input from overriding higher-priority instructions; apply output filtering to prevent disclosure of protected internal instructions; deploy prompt-injection detection and response guardrails. Junglewise finds every place you run it, drafts the upgrade and applies it once you approve.

**How is CVE-2026-126767 attacked?** It is attacked over the network, in a hosted service.

**Am I affected by CVE-2026-126767?** Check whether you run OpsTree REMS at an affected version. Junglewise does that check for you around the clock: connect your code, cloud and domains, and it matches new vulnerabilities like this one against what you actually run and tells you when one reaches you.

## How to fix


Implement strict separation between system, developer, retrieval, and user-controlled prompts; prevent user input from overriding higher-priority instructions; apply output filtering to prevent disclosure of protected internal instructions; deploy prompt-injection detection and response guardrails.

## Technical details

The RCA Chat endpoint (POST /api/v1/rca/chat) fails to properly isolate user-supplied input from system-level LLM prompts, allowing authenticated attackers to perform prompt injection attacks. By submitting crafted adversarial prompts, an attacker can cause the LLM to disclose internal instructions, workflow logic, and behavioral restrictions, which can be used to mount follow-up attacks.

## Affected products

- OpsTree REMS through 1.0.0

## Timeline

- 2026-09: disclosed: Disclosed by security researcher Neel Saurabh Kotnis
- 2026-11-11: advisory
- patched: Remediated by OpsTree Solutions; specific version not disclosed

## References

- [https://gist.github.com/neelkotnis/548bfde3e45bc3793f3f26e1f9a43efe](https://gist.github.com/neelkotnis/548bfde3e45bc3793f3f26e1f9a43efe)
- [https://opstree.com/rems-ai-powered-observability/](https://opstree.com/rems-ai-powered-observability/)

## Related threats

- [OpsTree REMS incorrect comment attribution](https://junglewise.ai/threats/cve-2026-126772-opstree-rems-through-1-0-0-contains-an-improper-attribution)
- [OpsTree REMS stored cross-site scripting in file preview](https://junglewise.ai/threats/cve-2026-126771-opstree-rems-through-1-0-0-contains-a-stored-cross-site)
- [OpsTree REMS server-side request forgery in database test connection](https://junglewise.ai/threats/cve-2026-126770-opstree-rems-through-1-0-0-contains-a-server-side-request)
- [OpsTree REMS improper authorization in incident creation](https://junglewise.ai/threats/cve-2026-126766-opstree-rems-through-1-0-0-contains-an-improper-authorization)