Server-Side Request Forgery in SciPhi-AI R2R API

A vulnerability was identified in SciPhi-AI R2R up to 3.6.6. This vulnerability affects unknown code of the file py/shared/abstractions/llm.py of the component Retrieval Completion API Endpoint. Such manipulation of the argument generation_config.api_base leads to server-side request forgery. The attack can be launched remotely. The exploit is publicly available and might be used. The vendor was contacted early about this disclosure but did not respond in any way.

Casky was already ahead

This CVE exploits attack patterns that Casky's 331matched skills already investigate — long before this vulnerability was disclosed. Claude's reasoning model maps these techniques to MITRE ATT&CK, so practitioners who ran these skills have already seen the threat behaviour in their findings.

Analysis

CVE-2026-105148 exposes a critical server-side request forgery (SSRF) vulnerability in SciPhi-AI R2R versions up to 3.6.6, specifically in the Retrieval Completion API Endpoint. By manipulating the generation_config.api_base argument, an unauthenticated remote attacker can force the server to make arbitrary HTTP requests to internal or external systems. This vulnerability is particularly dangerous because R2R is designed for retrieval-augmented generation (RAG) workflows, meaning affected systems often have privileged access to internal databases, APIs, and services. Organizations using vulnerable versions expose their infrastructure to unauthorized reconnaissance, data exfiltration, and potential lateral movement attacks. The lack of vendor responsiveness compounds the risk, leaving users without official guidance or patches.

Casky's 331 matching skills leverage Claude AI with extended reasoning to detect exploitation patterns associated with the two key MITRE ATT&CK techniques underlying this CVE: TA0001 (Initial Access) and TA0006 (Credential Access). Practitioners using Casky would identify suspicious API endpoint requests with modified base URLs pointing to internal IP ranges or localhost addresses, detect unusual outbound connections from the R2R service to unexpected destinations, and correlate API parameter tampering attempts with network traffic anomalies. The platform's skill mapping enables detection of reconnaissance probes testing API parameter boundaries, identification of credential harvesting attempts via SSRF chains to metadata services (AWS IMDSv1, Azure IMDS), and behavioral analysis of request patterns indicating automated exploitation. Practitioners would see findings highlighting: suspicious generation_config modifications, unexpected proxy or tunnel usage patterns, and outbound requests from the API endpoint to non-whitelisted targets—enabling rapid containment before attackers pivot deeper into infrastructure.

Live Threat IntelligenceComing soon

Casky Risk Score

—

EPSS Probability

—

Shodan Exposure

—

GreyNoise Signal

—

Composite risk scoring from EPSS, CISA KEV, Shodan, and GreyNoise — 21 security APIs correlated into a single Casky Risk Score. Coming in Casky Pro. Join early access →

331 Casky skills cover this attack pattern

These skills use Claude AI's reasoning model to surface findings in the same attack categories as CVE-2026-105148.

Abuse of Public-Facing API: Mobile API Abuse

mobile security · low

Run

abusing-dpapi-for-credential-access

red teaming · high

Run

abusing-shadow-credentials-for-privesc

red teaming · high

Run

MITRE ATT&CK Techniques

TA0001 TA0006

Investigate the attack patterns behind CVE-2026-105148

Casky has 331 skills that investigate the attack patterns behind CVE-2026-105148. Run one and get CVSS-scored findings in 3 minutes.

Run the skill that detects this →

X Instagram LinkedIn