Attack ChainAIDerived
Hypothetical attack flow derived from CVE metadata – not a confirmed exploit chain
-
AccessReach FinRobot agent input surface
-
DeliverySubmit malicious code-bearing prompt/data
-
ExploitReach create_file_with_code() unsanitized
-
ExecutionInjected code written and executed
-
ImpactArbitrary code execution on host
Vulnerability Assessment
| Weakness |
CWE
OWASP Top 10:2025
A05:2025 - Injection
ASVS 5.0
V1 - Encoding and Sanitization Code injection prevention, sandbox exec
MITRE ATT&CK
2 techniques across 2 tactics: Execution, Initial Access The technique list with ATT&CK links and the kill-chain view is part of the Basic plan and up. |
| Exploitation Conditions | Exploitation requires that attacker-controlled input reaches the CodingUtils.create_file_with_code() routine of FinRobot 1.0.0 - i.e., a deployment where the agent/orchestrator forwards untrusted text (a malicious prompt, document, or tool/data input) into this code-writing function without neutralization. The limiting factor is reachability: FinRobot is a Python agent framework typically run as a local/orchestrated process, so the 'remote unauthenticated' CVSS rating holds only when an integrator exposes the agent pipeline to external, untrusted input. No special product configuration flag is named in the description, but the practical precondition is an input path from an untrusted source to the vulnerable function; isolated/local-only usage with trusted input materially reduces real risk. |
| Impact | The input CVSS 3.1 vector (AV:N/AC:L/PR:N/UI:N, C:H/I:H/A:H → 9.8 Critical) describes an unauthenticated, network-reachable code-injection flaw (CWE-94) with full confidentiality/integrity/availability impact. However, the other signals pull strongly against treating this as an urgent, broadly-exploitable priority. EPSS is only 0.21% (10th percentile), it is NOT in CISA KEV (no confirmed active exploitation), and no SSVC decision data was provided. There is a clear tension to flag explicitly: the CVSS network/unauthenticated rating is almost certainly optimistic for this target - create_file_with_code() is a code-writing utility inside an AI financial-agent framework (FinRobot), and in typical deployments it is invoked by the local agent/orchestrator on LLM- or operator-supplied input rather than directly over an exposed, unauthenticated network socket. So 'network/unauthenticated' depends entirely on whether an integrator wires this function to untrusted external input; verify the actual exposure surface with the project. The practical priority is MODERATE: critical technical impact (code execution) but low exploitation probability, no KEV listing, and a small, niche deployment footprint. The GitHub issue (#102) and a Gist reference suggest exploit/PoC detail may exist and should be reviewed, but no PoC flag was confirmed in the input. |
| Exploit ScenarioDerived | An attacker who can influence the code/content argument passed to FinRobot's CodingUtils.create_file_with_code() - for example by feeding a malicious prompt or document into an agent workflow that forwards attacker-controlled text into this code-writing routine - injects Python/code that is written and subsequently executed, yielding arbitrary code execution in the host context (CWE-94). A GitHub issue (#102) and a public Gist are referenced and may contain PoC detail, but PoC availability is not confirmed in the provided data. The CVSS AC:L rating implies low complexity once attacker input reaches the vulnerable function. |
Get an e-mail when this CVE is exploited, patched or re-scored. Free account.
Alert meAnalysisAIv1 – Oct 05, 2026
Code injection in FinRobot 1.0.0's CodingUtils.create_file_with_code() allows unauthenticated remote attackers to execute arbitrary code when the agent pipeline is exposed to untrusted input, per the CVSS 3.1 vector (9.8 Critical, CWE-94). Exploitation requires that attacker-controlled text-such as a malicious prompt, document, or tool input-reaches the vulnerable function without neutralization; typical local or orchestrated deployments that only process trusted input materially reduce real risk. No public exploit identified at time of analysis, and EPSS is 0.21% (10th percentile), indicating low predicted exploitation activity.
RemediationAI
No vendor-released patch identified at time of analysis, so remediation must focus on reducing reachability and impact. Do not expose FinRobot's agent pipeline to untrusted external input; restrict or remove any interface that forwards user-supplied text, documents, or tool outputs to CodingUtils.create_file_with_code(). If the code-writing feature is not essential, disable or remove that function. For deployments that require it, enforce strict input validation and allowlist file paths/code templates, and run FinRobot in a low-privilege sandbox or container with a read-only filesystem and restricted network egress to contain any code execution. Trade-offs: sandboxing may break legitimate code-generation workflows, and disabling the feature removes functionality. Monitor https://github.com/AI4Finance-Foundation/FinRobot/issues/102 and https://nvd.nist.gov/vuln/detail/CVE-2026-51918 for patch updates.
Affected VersionsENISA
Affects your stack
Find out whether the stacks and servers you run are on an affected version.
Checking this CVE against your own stacks and servers is part of the Basic plan and up.
Technical ContextAI · show details
FinRobot is an open-source Python AI agent framework for financial analysis; the affected routine, CodingUtils.create_file_with_code(), dynamically writes files containing code generated from input. CWE-94 (Code Injection) indicates that the function fails to neutralize special elements (code syntax) in attacker-supplied data, so if that data is later evaluated or executed, arbitrary code runs with the privileges of the FinRobot process. The CVSS vector AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H rates this as network-reachable, low-complexity, unauthenticated, and requiring no user interaction, but the practical attack surface is gated by reachability: an integrator must expose the FinRobot agent/orchestrator to external untrusted input, such as via prompts, documents, or tool outputs, for the remote rating to apply. The CPE data is cpe:2.3:a:n/a:n/a:*:*:*:*:*:*:*:*, so no standardized vendor/product mapping is available beyond the FinRobot 1.0.0 identification.
Know when the next one hits your stack
- Watch products and get an email when something new is published
- Check this CVE against your own stack and servers
- Exploit and PoC references, threat intelligence from 22 sources
No credit card · 7-day full trial
Severity by source
AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H
Code injection yielding full code execution (C/I/A:H); AV:N/PR:N retained to match reachable-service deployments, though many installs are local-input-only, making AV:N the optimistic bound.
Primary rating from Vendor (mitre).
CVSS VectorVendor: mitre
Lifecycle Timeline
3DescriptionCVE.org
FinRobot 1.0.0 contains code injection in CodingUtils.create_file_with_code ().
Executive ViewAI
CVE-2026-51918 is a critical code-injection flaw in FinRobot 1.0.0 that lets crafted text reach the CodingUtils.create_file_with_code() function and cause arbitrary code execution inside the agent's own environment. Any team running FinRobot-based agents that accept external prompts, documents, or tool output is exposed, and because no vendor patch is available yet, security leadership should treat this as an open exposure requiring containment rather than waiting for a fix. The practical risk is concentrated in internet-facing or multi-tenant agent deployments; isolated instances processing only trusted internal input are materially less exposed.
No vendor-released patch identified at time of analysis. Interim mitigations: remove affected FinRobot 1.0.0 agents from internet and untrusted-network exposure; restrict or disable the CodingUtils.create_file_with_code() capability; sanitize and validate all prompts, documents, and tool inputs before they reach the agent pipeline; run agents in least-privilege, isolated containers with no production credentials, read-only filesystems, and blocked outbound egress; monitor for anomalous file writes and child processes; apply vendor-released fix as soon as it becomes available.
Threat Intelligence Overview
How this decision was reached
The engine reads each signal below and applies a fixed list of rules from top to bottom. The first rule that matches sets the recommended action – like a statement, every line is shown.
What every action level means
SSVC AssessmentENISA
Priority Score
Attack Technique
Code injection occurs when an application accepts user input and passes it directly into a language interpreter or evaluator without proper sanitization.
Remote Code Execution represents the critical moment when an attacker successfully runs arbitrary code on a target system without physical access.
EUVD-2026-91755
GHSA-xfxm-q6r8-5xm6
NVD
CVE.org
EPSS
Vulnrichment
Exploit-DB