Attack ChainAIDerived
Hypothetical attack flow derived from CVE metadata – not a confirmed exploit chain
-
AccessVictim runs agentflow serve on localhost
-
DeliveryAttacker hosts malicious web page
-
ExploitVictim browses malicious page in browser
-
ExecutionPage POSTs text/plain request to 127.0.0.1:8000/api/runs (no preflight)
-
PersistServer accepts non-JSON content type without validation
-
ImpactPipeline operation executed on local control plane
Vulnerability Assessment
| Weakness |
CWE
OWASP Top 10:2025
A07:2025 - Authentication Failures
ASVS 5.0
V6 - Authentication Origin validation
|
| Exploitation Conditions | Requires the victim to have AgentFlow installed and actively running |
| Impact | The CVSS 4.0 score of 4.8 (AV:L/AC:L/AT:N/PR:N/UI:P/VC:L/VI:L/VA:N) reflects that the vulnerable API is locally bound to 127.0.0.1 and exploitation requires the victim to interact with a browser (UI:P). The EPSS of 0.01% at the 0th percentile indicates near-zero observed exploitation probability, consistent with the tool's niche footprint. SSVC signals a PoC with non-automatable exploitation and partial technical impact - meaning mass scanning campaigns are not realistic but targeted abuse against a developer's workstation is plausible. No CISA KEV listing exists. The gap between the existing pipeline_path capability (arbitrary local .py execution) and the content-type bypass elevates the realistic impact above what raw CVSS C:L/I:L suggests: if pipeline_path was previously enabled or exposed to a developer with trust in the localhost API, the exploit chain approaches arbitrary local code execution. Absent that flag, impact is limited to pipeline submission and validation abuse. Overall, this is a genuine but low-urgency issue warranting patching rather than emergency response. |
| Exploit ScenarioDerived | An attacker hosts a web page containing JavaScript that POSTs a text/plain-typed body to http://127.0.0.1:8000/api/runs with a crafted pipeline payload. A developer who visits the page while |
Get an e-mail when this CVE is exploited, patched or re-scored. Free account.
Alert meAnalysisAIv1 – Jul 24, 2026
Content-type validation bypass in AgentFlow's localhost web API allows browser-driven cross-origin requests to reach the POST /api/runs and POST /api/runs/validate control plane endpoints without enforcement of application/json. Affected versions prior to commit 1667fa3 (CPE: cpe:2.3:a:berabuddies:agentflow:*) fail to reject non-JSON content types, enabling a malicious web page visited by a local AgentFlow user to issue CORS-simple requests that bypass the browser's same-origin enforcement and submit pipeline operations to the local agent runtime. No active exploitation is confirmed in CISA KEV; the SSVC framework records a PoC with partial technical impact and non-automatable exploitation.
RemediationAI
Update AgentFlow to a revision at or after commit 1667fa3 by pulling from the upstream repository (https://github.com/berabuddies/agentflow/commit/1667fa3). No tagged semantic release version is confirmed from available data - the fix exists as a commit and PR only. After updating, the API will return HTTP 415 for any POST to /api/runs or /api/runs/validate that does not carry Content-Type: application/json. Additionally, pipeline_path is now disabled by default in the web API; if your workflow requires it, enable it explicitly with AGENTFLOW_API_ALLOW_PIPELINE_PATH=1 only in trusted operator-controlled environments, as enabling it allows the API to load local .py files by path. As a pre-patch workaround, avoid running agentflow serve simultaneously with browsing untrusted websites, or use browser extensions that block outbound requests to localhost from web pages.
Affected VersionsENISA
Affected Packages
Source: OSV (Open Source Vulnerabilities)
Affects your stack
Affects 1 package in euvd
Checking this CVE against your own stacks and servers is part of the Basic plan and up.
Technical ContextAI · show details
AgentFlow (berabuddies, CPE: cpe:2.3:a:berabuddies:agentflow:*:*:*:*:*:*:*:*) is an AI pipeline orchestration tool that exposes a local HTTP API on 127.0.0.1:8000 when started with agentflow serve. CWE-346 (Origin Validation Error) is the root cause class: browsers classify POST requests carrying non-JSON content types such as text/plain as CORS 'simple requests,' which are dispatched without a preflight OPTIONS check. Because the pre-patch server did not enforce application/json on /api/runs and /api/runs/validate, a cross-origin page could submit pipeline payloads directly. Compounding this, the pipeline_path parameter was unrestricted by default, meaning a crafted request could instruct the server to load and execute an arbitrary local .py file by filesystem path. The fix (commit 1667fa3, PR #18) adds a _require_json_request() guard that returns HTTP 415 on non-JSON content types and gates pipeline_path behind an explicit AGENTFLOW_API_ALLOW_PIPELINE_PATH=1 environment variable.
Know when Agentflow gets its next vulnerability
- 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
CVSS:4.0/AV:L/AC:L/AT:N/PR:N/UI:P/VC:L/VI:L/VA:N/SC:N/SI:N/SA:N
Browser-driven cross-origin delivery makes AV:N appropriate; UI:R because victim must visit attacker page; no attacker privileges; limited C/I from pipeline abuse.
Primary rating from Vendor (VulnCheck).
CVSS VectorVendor: VulnCheck
Lifecycle Timeline
3DescriptionCVE.org
AgentFlow's local web API accepts non-JSON content types on POST /api/runs and POST /api/runs/validate endpoints without enforcing application/json validation, allowing attackers to bypass trust-boundary enforcement on sensitive operations. Attackers can exploit this content-type validation weakness through browser-driven or local cross-origin requests to abuse the localhost API and enable attack chains against the local control plane.
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
Information disclosure occurs when an application unintentionally exposes sensitive data that aids attackers in reconnaissance or directly compromises security.
Authentication bypass in Flowring Agentflow workflow system allows unauthenticated remote attackers to exploit specific
Agentflow from Flowring Technology has an Account Lockout Bypass vulnerability, allowing unauthenticated remote attacker
Missing authentication in Flowring Agentflow allows unauthenticated attackers to read, modify, and delete data. Second a
The file upload function of Agentflow BPM has insufficient filtering for special characters in URLs. Rated critical seve
Agentflow versions up to - is affected by unrestricted upload of file with dangerous type (CVSS 8.8).
Agentflow BPM enterprise management system has improper authentication. Rated high severity (CVSS 8.8), this vulnerabili
Arbitrary code execution in berabuddies AgentFlow (all versions prior to fix commit 1667fa3) allows attackers to run loc
Agentflow BPM file download function has a path traversal vulnerability. Rated high severity (CVSS 7.5), this vulnerabil
Reflected XSS in AgentFlow enables unauthenticated attackers to inject malicious JavaScript that executes in victims' br
Authenticated attackers can inject malicious JavaScript into Flowring's AgentFlow platform that persists and executes in
EUVD-2026-26278
NVD
CVE.org
EPSS
Vulnrichment
GitHub
VulDB
Exploit-DB