CVE-2026-84888: RightNow-AI OpenFang RCE Vulnerability

CVE-2026-84888 is a remote code execution flaw in RightNow-AI OpenFang up to version 0.6.9 affecting the shell_exec function. Attackers can exploit this remotely to cause uncontrolled memory allocation. This article covers technical details, affected versions, impact assessment, and mitigation strategies.

CVE-2026-84888 Overview

CVE-2026-84888 identifies an uncontrolled resource consumption weakness in RightNow-AI OpenFang versions up to 0.6.9. The flaw resides in the shell_exec function within crates/openfang-runtime/src/tool_runner.rs. Manipulating input to this function triggers uncontrolled memory allocation, enabling a remote denial-of-service condition. Public exploit details have been released, and the vendor did not respond to disclosure attempts. The vulnerability is tracked under [CWE-400] (Uncontrolled Resource Consumption).

Critical Impact
Remote attackers with low privileges can trigger uncontrolled memory allocation in the OpenFang runtime, exhausting host memory and producing a denial-of-service condition against the affected process.

Affected Products

RightNow-AI OpenFang versions up to and including 0.6.9
Component: crates/openfang-runtime/src/tool_runner.rs
Function: shell_exec

Discovery Timeline

2026-09-03 - CVE-2026-84888 published to NVD
2026-09-03 - Last updated in NVD database

Technical Details for CVE-2026-84888

Vulnerability Analysis

The vulnerability affects the shell_exec tool runner in OpenFang, an AI runtime component distributed by RightNow-AI. When shell_exec executes a subprocess, it captures the process output into a buffer without enforcing an upper bound on allocation size. An attacker who can influence the command or its output stream can force the runtime to allocate memory proportional to the output produced by the invoked shell process.

Because the runtime lacks a size ceiling, sustained or large output causes the host to allocate memory until the process is terminated by the operating system. This is a classic uncontrolled resource consumption pattern classified under [CWE-400]. The condition is reachable over the network with low privileges and does not require user interaction.

Impact is limited to availability. The vulnerability does not disclose data, alter integrity, or grant code execution. However, the GitHub CVE Analysis Report documents a reproducible exploit path.

Root Cause

The root cause is missing output-size enforcement in the shell_exec implementation. The runtime reads subprocess stdout and stderr into an in-memory buffer without applying byte limits, streaming to disk, or truncating oversized responses. Any command whose output grows unbounded will drive Rust's allocator to consume all available memory.

Attack Vector

An authenticated remote client submits a tool invocation that reaches shell_exec and produces a large or unbounded output stream. Commands such as those reading from /dev/zero, generating recursive listings, or looping infinitely are sufficient to trigger the condition. The runtime process then exhausts memory and is terminated, disrupting availability of the OpenFang service.

No verified proof-of-concept code has been reproduced in this article. Refer to the VulDB entry for CVE-2026-84888 and the GitHub CVE Analysis Report for technical details.

Detection Methods for CVE-2026-84888

Indicators of Compromise

Sudden growth in resident memory of the OpenFang runtime process followed by OOMKilled events or process crashes.
Repeated tool invocations targeting shell_exec with commands that generate large or streaming output.
Kernel oom-killer log entries referencing the openfang-runtime binary or its worker processes.

Detection Strategies

Instrument the runtime host with resource-usage telemetry and alert on rapid memory pressure attributed to OpenFang processes.
Enable application-level logging of shell_exec tool calls, including command strings, output byte counts, and execution duration.
Correlate authentication events for the OpenFang API with anomalous tool-call volume from the same principal.

Monitoring Recommendations

Track process-level RSS and VSZ metrics for OpenFang workers and set thresholds tied to host capacity.
Monitor container restart counts and OOM events in orchestrators such as Kubernetes for pods running OpenFang.
Capture and retain tool-runner audit logs in a centralized data lake for retrospective analysis of abusive command patterns.

How to Mitigate CVE-2026-84888

Immediate Actions Required

Restrict network exposure of the OpenFang runtime to trusted callers only, using network policies or a reverse proxy with authentication.
Enforce per-process memory limits on the runtime using cgroups, container --memory flags, or systemdMemoryMax directives.
Rate-limit and audit calls to the shell_exec tool endpoint, and revoke credentials for principals exhibiting abusive patterns.

Patch Information

No vendor patch is referenced in the published advisory. The vendor did not respond to the disclosure. Operators should monitor the RightNow-AI OpenFang project references and the VulDB entry for future fix availability. Until a patch is released, apply the workarounds below.

Workarounds

Disable or remove the shell_exec tool from the runtime tool registry if the deployment does not require shell execution.
Wrap or replace shell_exec with a bounded implementation that caps captured output bytes and enforces an execution timeout.
Run OpenFang inside a container or sandbox with strict memory and CPU limits so that a single abusive invocation cannot destabilize the host.
Require least-privilege API tokens for tool invocation and log every shell_exec call for post-incident review.

bash

# Example: enforce a hard memory ceiling on the OpenFang runtime container
docker run --name openfang \
  --memory=512m \
  --memory-swap=512m \
  --pids-limit=128 \
  --read-only \
  rightnow-ai/openfang:latest