CVE-2026-15618: Mosaxiv Clawlet RCE Vulnerability

CVE-2026-15618 Overview

CVE-2026-15618 is a protection mechanism failure vulnerability in the mosaxiv clawlet project through version 0.2.10. The flaw resides in the guardExecCommand function within tools/tool_exec.go, part of the exec Safety Guard component. Manipulation of input to this function causes the safety guard to fail, allowing operations that the guard is intended to block. The issue is remotely triggerable and requires user interaction. A public exploit has been released, and the maintainer closed the associated GitHub issue with the label "not planned," indicating no vendor fix is expected.

Critical Impact
Attackers can bypass the exec Safety Guard in tools/tool_exec.go, undermining protections around command execution. No official patch has been released.

Affected Products

mosaxiv clawlet versions up to and including 0.2.10
Component: exec Safety Guard (tools/tool_exec.go)
Function: guardExecCommand

Discovery Timeline

2026-07-14 - CVE-2026-15618 published to the National Vulnerability Database (NVD)
2026-07-14 - Last updated in NVD database

Technical Details for CVE-2026-15618

Vulnerability Analysis

The vulnerability is classified under [CWE-693] Protection Mechanism Failure. The guardExecCommand function in tools/tool_exec.go is designed to validate or restrict commands before they are executed by the clawlet tool. When specifically crafted inputs are supplied, the guard fails to enforce its intended restrictions, allowing operations that should have been blocked.

The attack path is remote, requires low complexity, and depends on user interaction to trigger. Confidentiality, integrity, and availability impacts are limited in scope. According to the vendor response, the reported GitHub issue was closed as "not planned," so an upstream code fix should not be expected.

Root Cause

The root cause is insufficient enforcement logic inside guardExecCommand. The function does not adequately validate command inputs against its safety policy, permitting bypass of the exec Safety Guard. Because the guard is the sole control preventing unsafe command execution, its failure removes the intended protection boundary between untrusted input and the command execution surface.

Attack Vector

Exploitation occurs over the network and requires a user to perform an action that submits crafted input to the vulnerable function. Once the guard is bypassed, downstream execution paths in the tool operate without the intended restrictions. A public exploit is available, which lowers the effort required for attackers to reproduce the bypass.

No verified code example is available. Refer to the GitHub Clawlet Issue #12 and VulDB CVE-2026-15618 entry for technical detail on the reported bypass.

Detection Methods for CVE-2026-15618

Indicators of Compromise

Unexpected process executions originating from clawlet or its tool_exec.go code paths.
Command executions that should have been rejected by the exec Safety Guard appearing in audit logs.
Presence of clawlet binaries or dependencies at version 0.2.10 or earlier in the environment.

Detection Strategies

Inventory Go projects and developer tooling to identify installations of mosaxiv clawlet ≤ 0.2.10.
Instrument or wrap invocations of guardExecCommand to log both the input and the guard decision for offline review.
Correlate child process creation events from clawlet with the commands the guard reports as denied to identify divergence.

Monitoring Recommendations

Enable process creation auditing on hosts running clawlet and forward events to a centralized log platform.
Alert on clawlet spawning shells, interpreters, or network utilities that are not part of expected workflows.
Track file integrity for tools/tool_exec.go and related binaries to detect unauthorized modifications or patches.

How to Mitigate CVE-2026-15618

Immediate Actions Required

Restrict use of mosaxiv clawlet ≤ 0.2.10 to trusted inputs and isolated environments until a fix is available.
Run clawlet under a least-privilege user account with no ability to execute sensitive commands.
Block or monitor untrusted user interaction paths that can reach the guardExecCommand function.

Patch Information

No official patch has been released. The maintainer closed GitHub Clawlet Issue #12 with the label "not planned." Consult the GitHub Clawlet Repository for any future updates and consider maintaining a private fork with additional validation in guardExecCommand if continued use is required.

Workarounds

Sandbox clawlet execution using containers, seccomp profiles, or AppArmor/SELinux policies that restrict allowable commands.
Implement an external allowlist proxy in front of clawlet that validates commands independently of guardExecCommand.
Remove clawlet from production or CI/CD pipelines that process untrusted input if compensating controls cannot be applied.

bash

# Configuration example: restrict clawlet with a minimal seccomp/AppArmor-style wrapper
# 1. Run clawlet as an unprivileged user
useradd -r -s /usr/sbin/nologin clawlet-svc

# 2. Constrain execution with systemd hardening
# /etc/systemd/system/clawlet.service
# [Service]
# User=clawlet-svc
# NoNewPrivileges=true
# ProtectSystem=strict
# ProtectHome=true
# PrivateTmp=true
# SystemCallFilter=@system-service
# RestrictSUIDSGID=true
# CapabilityBoundingSet=