CVE-2026-16125: zevorn rt-claw SSRF Vulnerability

CVE-2026-16125 Overview

CVE-2026-16125 is a Server-Side Request Forgery (SSRF) vulnerability affecting zevorn rt-claw through version 0.2.0. The flaw resides in the claw_net_get and claw_net_post functions within claw/services/tools/net.c, part of the http_request component. Attackers can manipulate the url argument to coerce the server into issuing arbitrary HTTP requests to unintended destinations. The vulnerability is exploitable remotely without authentication or user interaction. A public exploit has been disclosed, and the project maintainers have not yet responded to the issue report filed against the repository.

Critical Impact
Remote attackers can abuse the url parameter passed to claw_net_get/claw_net_post to force the server to send crafted HTTP requests to internal or external resources, enabling reconnaissance and interaction with services otherwise unreachable from the network edge.

Affected Products

zevorn rt-claw versions up to and including 0.2.0
Component: http_request (claw/services/tools/net.c)
Affected functions: claw_net_get, claw_net_post

Discovery Timeline

2026-07-18 - CVE-2026-16125 published to NVD
2026-07-21 - Last updated in NVD database

Technical Details for CVE-2026-16125

Vulnerability Analysis

The vulnerability is classified under CWE-918: Server-Side Request Forgery. The claw_net_get and claw_net_post functions in claw/services/tools/net.c accept a url argument that is passed directly into the outbound HTTP request logic without adequate validation against destination hosts, schemes, or address ranges. An attacker can supply a URL pointing to internal-only services, loopback interfaces, cloud metadata endpoints, or other restricted destinations. Because the request originates from the server, it inherits the server's network position and any implicit trust relationships. Public disclosure of the flaw increases the likelihood of opportunistic exploitation against exposed instances.

Root Cause

The root cause is missing validation of the user-controlled url parameter inside claw_net_get and claw_net_post. Neither function enforces an allow-list of hosts, restricts the URL scheme, nor blocks requests to internal IP ranges before initiating the outbound HTTP request. This lets attacker-supplied input dictate the destination of a server-originated network call.

Attack Vector

Exploitation occurs over the network with no authentication or user interaction required. An attacker submits a crafted url value to any endpoint or interface that routes into claw_net_get or claw_net_post. The server then issues the attacker-directed HTTP request, and the response or side effects can be observed depending on how rt-claw returns network data to callers. Refer to the GitHub issue discussion and the VulDB vulnerability details for further technical context.

No verified proof-of-concept code is published in a structured form. The vulnerability mechanism is described in prose above; see the GitHub PoC repository for source-level context.

Detection Methods for CVE-2026-16125

Indicators of Compromise

Outbound HTTP requests from rt-claw processes to internal IP ranges such as 127.0.0.0/8, 10.0.0.0/8, 172.16.0.0/12, or 192.168.0.0/16.
Requests from the server to cloud metadata endpoints, notably 169.254.169.254.
Unusual URL schemes or hostnames appearing in application logs tied to claw_net_get or claw_net_post invocations.

Detection Strategies

Instrument logging around claw_net_get and claw_net_post to capture every resolved destination host and port before the request is issued.
Compare outbound destinations against an expected allow-list and alert on deviations.
Correlate application-layer request logs with network egress telemetry to identify server-initiated traffic to non-business destinations.

Monitoring Recommendations

Monitor egress traffic from hosts running rt-claw for connections to RFC1918 addresses, loopback ranges, or link-local metadata services.
Alert on repeated failed outbound requests originating from the application service account, which can indicate SSRF probing.
Track error rates and response-size anomalies from the http_request component that could reveal reconnaissance activity.

How to Mitigate CVE-2026-16125

Immediate Actions Required

Restrict network egress from hosts running rt-claw so that only required destinations are reachable.
Block outbound access from the application to cloud metadata endpoints and internal management interfaces at the network layer.
Audit any code paths that expose claw_net_get or claw_net_post to user-controlled input and disable them where feasible until a fix is available.

Patch Information

No vendor patch is currently available. According to the disclosure, the project was informed early through an issue report but has not responded. Track the GitHub repository and the issue thread for updates, and consult the VulDB CVE listing for status changes.

Workarounds

Add input validation in front of claw_net_get/claw_net_post to reject URLs whose resolved IP falls within loopback, private, link-local, or reserved ranges.
Enforce a scheme allow-list limited to http and https, and reject file://, gopher://, and other unexpected schemes.
Route outbound HTTP calls through a filtering proxy that enforces destination allow-listing at the network boundary.
Apply DNS pinning or resolve-once-then-connect patterns to prevent DNS rebinding attacks against the validator.

bash

# Example egress firewall rule blocking access to cloud metadata endpoints
iptables -A OUTPUT -m owner --uid-owner rt-claw -d 169.254.169.254 -j REJECT
iptables -A OUTPUT -m owner --uid-owner rt-claw -d 127.0.0.0/8 -j REJECT
iptables -A OUTPUT -m owner --uid-owner rt-claw -d 10.0.0.0/8 -j REJECT
iptables -A OUTPUT -m owner --uid-owner rt-claw -d 172.16.0.0/12 -j REJECT
iptables -A OUTPUT -m owner --uid-owner rt-claw -d 192.168.0.0/16 -j REJECT