GHSA-V8HW-MH8C-JXFC
Vulnerability from github – Published: 2026-03-26 18:31 – Updated: 2026-06-06 00:56Description
1. Summary
The Agentic Assistant feature in Langflow executes LLM-generated Python code during its validation phase. Although this phase appears intended to validate generated component code, the implementation reaches dynamic execution sinks and instantiates the generated class server-side.
In deployments where an attacker can access the Agentic Assistant feature and influence the model output, this can result in arbitrary server-side Python execution.
2. Description
2.1 Intended Functionality
The Agentic Assistant endpoints are designed to help users generate and validate components for a flow. Users can submit requests to the assistant, which returns candidate component code for further processing.
A reasonable security expectation is that validation should treat model output as untrusted text and perform only static or side-effect-free checks.
The externally reachable endpoints are:
The request model accepts attacker-influenceable fields such as input_value, flow_id, provider, model_name, session_id, and max_retries:
2.2 Root Cause
In the affected code path, Langflow processes model output through the following chain:
/assist
→ execute_flow_with_validation()
→ execute_flow_file()
→ LLM returns component code
→ extract_component_code()
→ validate_component_code()
→ create_class()
→ generated class is instantiated
The assistant service reaches the validation path here:
The code extraction step occurs here:
The validation entry point is here:
The issue is that this validation path is not purely static. It ultimately invokes create_class() in lfx.custom.validate, where Python code is dynamically executed via exec(...), including both global-scope preparation and class construction.
As a result, LLM-generated code is treated as executable Python rather than inert data. This means the “validation” step crosses a trust boundary and becomes an execution sink.
The streaming path can also reach this sink when the request is classified into the component-generation branch:
3. Proof of Concept (PoC)
- Send a request to the Agentic Assistant endpoint.
- Provide input that causes the model to return malicious component code.
- The returned code reaches the validation path.
- During validation, the server dynamically executes the generated Python.
- Arbitrary server-side code execution occurs.
4. Impact
- Attackers who can access the Agentic Assistant feature and influence model output may execute arbitrary Python code on the server.
-
This can lead to:
-
OS command execution
- file read/write
- credential or secret disclosure
- full compromise of the Langflow process
5. Exploitability Notes
This issue is most accurately described as an authenticated or feature-reachable code execution vulnerability, rather than an unconditional unauthenticated remote attack.
Severity depends on deployment model:
- In local-only, single-user development setups, the issue may be limited to self-exposure by the operator.
- In shared, team, or internet-exposed deployments, it may be exploitable by other users or attackers who can reach the assistant feature.
The assistant feature depends on an active user context:
Authentication sources include bearer token, cookie, or API key:
Default deployment settings may widen exposure, including AUTO_LOGIN=true and the /api/v1/auto_login endpoint:
6. Patch Recommendation
- Remove all dynamic execution from the validation path.
- Ensure validation is strictly static and side-effect-free.
- Treat all LLM output as untrusted input.
- If code generation must be supported, require explicit approval and run it in a hardened sandbox isolated from the main server process.
Discovered by: @kexinoh (https://github.com/kexinoh, works at Tencent Zhuque Lab)
{
"affected": [
{
"database_specific": {
"last_known_affected_version_range": "\u003c= 1.8.1"
},
"package": {
"ecosystem": "PyPI",
"name": "langflow"
},
"ranges": [
{
"events": [
{
"introduced": "0"
},
{
"fixed": "1.9.0"
}
],
"type": "ECOSYSTEM"
}
]
}
],
"aliases": [
"CVE-2026-33873"
],
"database_specific": {
"cwe_ids": [
"CWE-94"
],
"github_reviewed": true,
"github_reviewed_at": "2026-03-26T18:31:36Z",
"nvd_published_at": "2026-03-27T21:17:23Z",
"severity": "CRITICAL"
},
"details": "## Description\n\n### 1. Summary\n\nThe Agentic Assistant feature in Langflow executes LLM-generated Python code during its **validation** phase. Although this phase appears intended to validate generated component code, the implementation reaches dynamic execution sinks and instantiates the generated class server-side.\n\nIn deployments where an attacker can access the Agentic Assistant feature and influence the model output, this can result in arbitrary server-side Python execution.\n\n### 2. Description\n\n#### 2.1 Intended Functionality\n\nThe Agentic Assistant endpoints are designed to help users generate and validate components for a flow. Users can submit requests to the assistant, which returns candidate component code for further processing.\n\nA reasonable security expectation is that validation should treat model output as **untrusted text** and perform only static or side-effect-free checks.\n\nThe externally reachable endpoints are:\n\n[https://github.com/langflow-ai/langflow/blob/f7f4d1e70ba5eecd18162ec96f3571c2cfbcd1fc/src/backend/base/langflow/agentic/api/router.py#L252-L297](https://github.com/langflow-ai/langflow/blob/f7f4d1e70ba5eecd18162ec96f3571c2cfbcd1fc/src/backend/base/langflow/agentic/api/router.py#L252-L297)\n\nThe request model accepts attacker-influenceable fields such as `input_value`, `flow_id`, `provider`, `model_name`, `session_id`, and `max_retries`:\n\n[https://github.com/langflow-ai/langflow/blob/f7f4d1e70ba5eecd18162ec96f3571c2cfbcd1fc/src/backend/base/langflow/agentic/api/schemas.py#L20-L31](https://github.com/langflow-ai/langflow/blob/f7f4d1e70ba5eecd18162ec96f3571c2cfbcd1fc/src/backend/base/langflow/agentic/api/schemas.py#L20-L31)\n\n#### 2.2 Root Cause\n\nIn the affected code path, Langflow processes model output through the following chain:\n\n`/assist`\n\u2192 `execute_flow_with_validation()`\n\u2192 `execute_flow_file()`\n\u2192 LLM returns component code\n\u2192 `extract_component_code()`\n\u2192 `validate_component_code()`\n\u2192 `create_class()`\n\u2192 generated class is instantiated\n\nThe assistant service reaches the validation path here:\n\n[https://github.com/langflow-ai/langflow/blob/f7f4d1e70ba5eecd18162ec96f3571c2cfbcd1fc/src/backend/base/langflow/agentic/services/assistant_service.py#L58-L79](https://github.com/langflow-ai/langflow/blob/f7f4d1e70ba5eecd18162ec96f3571c2cfbcd1fc/src/backend/base/langflow/agentic/services/assistant_service.py#L58-L79)\n\nThe code extraction step occurs here:\n\n[https://github.com/langflow-ai/langflow/blob/f7f4d1e70ba5eecd18162ec96f3571c2cfbcd1fc/src/backend/base/langflow/agentic/helpers/code_extraction.py#L11-L53](https://github.com/langflow-ai/langflow/blob/f7f4d1e70ba5eecd18162ec96f3571c2cfbcd1fc/src/backend/base/langflow/agentic/helpers/code_extraction.py#L11-L53)\n\nThe validation entry point is here:\n\n[https://github.com/langflow-ai/langflow/blob/f7f4d1e70ba5eecd18162ec96f3571c2cfbcd1fc/src/backend/base/langflow/agentic/helpers/validation.py#L27-L47](https://github.com/langflow-ai/langflow/blob/f7f4d1e70ba5eecd18162ec96f3571c2cfbcd1fc/src/backend/base/langflow/agentic/helpers/validation.py#L27-L47)\n\nThe issue is that this validation path is not purely static. It ultimately invokes `create_class()` in `lfx.custom.validate`, where Python code is dynamically executed via `exec(...)`, including both global-scope preparation and class construction.\n\n[https://github.com/langflow-ai/langflow/blob/f7f4d1e70ba5eecd18162ec96f3571c2cfbcd1fc/src/lfx/src/lfx/custom/validate.py#L241-L272](https://github.com/langflow-ai/langflow/blob/f7f4d1e70ba5eecd18162ec96f3571c2cfbcd1fc/src/lfx/src/lfx/custom/validate.py#L241-L272)\n\n[https://github.com/langflow-ai/langflow/blob/f7f4d1e70ba5eecd18162ec96f3571c2cfbcd1fc/src/lfx/src/lfx/custom/validate.py#L394-L399](https://github.com/langflow-ai/langflow/blob/f7f4d1e70ba5eecd18162ec96f3571c2cfbcd1fc/src/lfx/src/lfx/custom/validate.py#L394-L399)\n\n[https://github.com/langflow-ai/langflow/blob/f7f4d1e70ba5eecd18162ec96f3571c2cfbcd1fc/src/lfx/src/lfx/custom/validate.py#L441-L443](https://github.com/langflow-ai/langflow/blob/f7f4d1e70ba5eecd18162ec96f3571c2cfbcd1fc/src/lfx/src/lfx/custom/validate.py#L441-L443)\n\nAs a result, LLM-generated code is treated as executable Python rather than inert data. This means the \u201cvalidation\u201d step crosses a trust boundary and becomes an execution sink.\n\nThe streaming path can also reach this sink when the request is classified into the component-generation branch:\n\n[https://github.com/langflow-ai/langflow/blob/f7f4d1e70ba5eecd18162ec96f3571c2cfbcd1fc/src/backend/base/langflow/agentic/services/assistant_service.py#L142-L156](https://github.com/langflow-ai/langflow/blob/f7f4d1e70ba5eecd18162ec96f3571c2cfbcd1fc/src/backend/base/langflow/agentic/services/assistant_service.py#L142-L156)\n\n[https://github.com/langflow-ai/langflow/blob/f7f4d1e70ba5eecd18162ec96f3571c2cfbcd1fc/src/backend/base/langflow/agentic/services/assistant_service.py#L259-L300](https://github.com/langflow-ai/langflow/blob/f7f4d1e70ba5eecd18162ec96f3571c2cfbcd1fc/src/backend/base/langflow/agentic/services/assistant_service.py#L259-L300)\n\n### 3. Proof of Concept (PoC)\n\n1. Send a request to the Agentic Assistant endpoint.\n2. Provide input that causes the model to return malicious component code.\n3. The returned code reaches the validation path.\n4. During validation, the server dynamically executes the generated Python.\n5. Arbitrary server-side code execution occurs.\n\n### 4. Impact\n\n* Attackers who can access the Agentic Assistant feature and influence model output may execute arbitrary Python code on the server.\n* This can lead to:\n\n * OS command execution\n * file read/write\n * credential or secret disclosure\n * full compromise of the Langflow process\n\n### 5. Exploitability Notes\n\nThis issue is most accurately described as an **authenticated or feature-reachable code execution vulnerability**, rather than an unconditional unauthenticated remote attack.\n\nSeverity depends on deployment model:\n\n* In **local-only, single-user development setups**, the issue may be limited to self-exposure by the operator.\n* In **shared, team, or internet-exposed deployments**, it may be exploitable by other users or attackers who can reach the assistant feature.\n\nThe assistant feature depends on an active user context:\n\n[https://github.com/langflow-ai/langflow/blob/f7f4d1e70ba5eecd18162ec96f3571c2cfbcd1fc/src/backend/base/langflow/api/utils/core.py#L38](https://github.com/langflow-ai/langflow/blob/f7f4d1e70ba5eecd18162ec96f3571c2cfbcd1fc/src/backend/base/langflow/api/utils/core.py#L38)\n\nAuthentication sources include bearer token, cookie, or API key:\n\n[https://github.com/langflow-ai/langflow/blob/f7f4d1e70ba5eecd18162ec96f3571c2cfbcd1fc/src/backend/base/langflow/services/auth/utils.py#L39-L53](https://github.com/langflow-ai/langflow/blob/f7f4d1e70ba5eecd18162ec96f3571c2cfbcd1fc/src/backend/base/langflow/services/auth/utils.py#L39-L53)\n\n[https://github.com/langflow-ai/langflow/blob/f7f4d1e70ba5eecd18162ec96f3571c2cfbcd1fc/src/backend/base/langflow/services/auth/utils.py#L156-L163](https://github.com/langflow-ai/langflow/blob/f7f4d1e70ba5eecd18162ec96f3571c2cfbcd1fc/src/backend/base/langflow/services/auth/utils.py#L156-L163)\n\nDefault deployment settings may widen exposure, including `AUTO_LOGIN=true` and the `/api/v1/auto_login` endpoint:\n\n[https://github.com/langflow-ai/langflow/blob/f7f4d1e70ba5eecd18162ec96f3571c2cfbcd1fc/src/lfx/src/lfx/services/settings/auth.py#L71-L87](https://github.com/langflow-ai/langflow/blob/f7f4d1e70ba5eecd18162ec96f3571c2cfbcd1fc/src/lfx/src/lfx/services/settings/auth.py#L71-L87)\n\n[https://github.com/langflow-ai/langflow/blob/f7f4d1e70ba5eecd18162ec96f3571c2cfbcd1fc/src/backend/base/langflow/api/v1/login.py#L96-L135](https://github.com/langflow-ai/langflow/blob/f7f4d1e70ba5eecd18162ec96f3571c2cfbcd1fc/src/backend/base/langflow/api/v1/login.py#L96-L135)\n\n### 6. Patch Recommendation\n\n* Remove all dynamic execution from the validation path.\n* Ensure validation is strictly static and side-effect-free.\n* Treat all LLM output as untrusted input.\n* If code generation must be supported, require explicit approval and run it in a hardened sandbox isolated from the main server process.\n\nDiscovered by: @kexinoh ([https://github.com/kexinoh](https://github.com/kexinoh), works at Tencent Zhuque Lab)",
"id": "GHSA-v8hw-mh8c-jxfc",
"modified": "2026-06-06T00:56:57Z",
"published": "2026-03-26T18:31:36Z",
"references": [
{
"type": "WEB",
"url": "https://github.com/langflow-ai/langflow/security/advisories/GHSA-v8hw-mh8c-jxfc"
},
{
"type": "ADVISORY",
"url": "https://nvd.nist.gov/vuln/detail/CVE-2026-33873"
},
{
"type": "WEB",
"url": "https://github.com/pypa/advisory-database/tree/main/vulns/langflow/PYSEC-2026-82.yaml"
},
{
"type": "WEB",
"url": "https://github.com/langflow-ai/langflow/blob/f7f4d1e70ba5eecd18162ec96f3571c2cfbcd1fc/src/lfx/src/lfx/services/settings/auth.py#L71-L87"
},
{
"type": "WEB",
"url": "https://github.com/langflow-ai/langflow/blob/f7f4d1e70ba5eecd18162ec96f3571c2cfbcd1fc/src/lfx/src/lfx/custom/validate.py#L441-L443"
},
{
"type": "WEB",
"url": "https://github.com/langflow-ai/langflow/blob/f7f4d1e70ba5eecd18162ec96f3571c2cfbcd1fc/src/lfx/src/lfx/custom/validate.py#L394-L399"
},
{
"type": "WEB",
"url": "https://github.com/langflow-ai/langflow/blob/f7f4d1e70ba5eecd18162ec96f3571c2cfbcd1fc/src/lfx/src/lfx/custom/validate.py#L241-L272"
},
{
"type": "WEB",
"url": "https://github.com/langflow-ai/langflow/blob/f7f4d1e70ba5eecd18162ec96f3571c2cfbcd1fc/src/backend/base/langflow/services/auth/utils.py#L39-L53"
},
{
"type": "WEB",
"url": "https://github.com/langflow-ai/langflow/blob/f7f4d1e70ba5eecd18162ec96f3571c2cfbcd1fc/src/backend/base/langflow/services/auth/utils.py#L156-L163"
},
{
"type": "WEB",
"url": "https://github.com/langflow-ai/langflow/blob/f7f4d1e70ba5eecd18162ec96f3571c2cfbcd1fc/src/backend/base/langflow/api/v1/login.py#L96-L135"
},
{
"type": "WEB",
"url": "https://github.com/langflow-ai/langflow/blob/f7f4d1e70ba5eecd18162ec96f3571c2cfbcd1fc/src/backend/base/langflow/api/utils/core.py#L38"
},
{
"type": "WEB",
"url": "https://github.com/langflow-ai/langflow/blob/f7f4d1e70ba5eecd18162ec96f3571c2cfbcd1fc/src/backend/base/langflow/agentic/services/assistant_service.py#L58-L79"
},
{
"type": "WEB",
"url": "https://github.com/langflow-ai/langflow/blob/f7f4d1e70ba5eecd18162ec96f3571c2cfbcd1fc/src/backend/base/langflow/agentic/services/assistant_service.py#L259-L300"
},
{
"type": "WEB",
"url": "https://github.com/langflow-ai/langflow/blob/f7f4d1e70ba5eecd18162ec96f3571c2cfbcd1fc/src/backend/base/langflow/agentic/services/assistant_service.py#L142-L156"
},
{
"type": "WEB",
"url": "https://github.com/langflow-ai/langflow/blob/f7f4d1e70ba5eecd18162ec96f3571c2cfbcd1fc/src/backend/base/langflow/agentic/helpers/validation.py#L27-L47"
},
{
"type": "WEB",
"url": "https://github.com/langflow-ai/langflow/blob/f7f4d1e70ba5eecd18162ec96f3571c2cfbcd1fc/src/backend/base/langflow/agentic/helpers/code_extraction.py#L11-L53"
},
{
"type": "WEB",
"url": "https://github.com/langflow-ai/langflow/blob/f7f4d1e70ba5eecd18162ec96f3571c2cfbcd1fc/src/backend/base/langflow/agentic/api/schemas.py#L20-L31"
},
{
"type": "WEB",
"url": "https://github.com/langflow-ai/langflow/blob/f7f4d1e70ba5eecd18162ec96f3571c2cfbcd1fc/src/backend/base/langflow/agentic/api/router.py#L252-L297"
},
{
"type": "PACKAGE",
"url": "https://github.com/langflow-ai/langflow"
}
],
"schema_version": "1.4.0",
"severity": [
{
"score": "CVSS:4.0/AV:N/AC:L/AT:N/PR:L/UI:N/VC:H/VI:H/VA:N/SC:H/SI:H/SA:N",
"type": "CVSS_V4"
}
],
"summary": "Langflow has Authenticated Code Execution in Agentic Assistant Validation"
}
Drag nodes to arrange them. Double-click a related vulnerability to open it, or a source to visit its link. The Physics button on the left tunes the layout and the simulation; the legend at the bottom left hides or shows nodes by kind and by sighting type.
Sightings
| Author | Source | Type | Date | Other |
|---|
Nomenclature
- Seen: The vulnerability was mentioned, discussed, or observed by the user.
- Confirmed: The vulnerability has been validated from an analyst's perspective.
- Published Proof of Concept: A public proof of concept is available for this vulnerability.
- Exploited: The vulnerability was observed as exploited by the user who reported the sighting.
- Patched: The vulnerability was observed as successfully patched by the user who reported the sighting.
- Not exploited: The vulnerability was not observed as exploited by the user who reported the sighting.
- Not confirmed: The user expressed doubt about the validity of the vulnerability.
- Not patched: The vulnerability was not observed as successfully patched by the user who reported the sighting.
Vulnerability timeline
Sightings and recorded dates, grouped by day · UTC
How durations are calculated
Following the VTEM framework (Black Hat USA 2015), time to patch starts when the vendor or a coordinator was informed, total time to patch at discovery, and disclosure to patch at public disclosure; vendor response time runs from the report to its acknowledgement. Time to exploit and time unpatched start at the earliest discovery or report, falling back to public disclosure. Unpatched time ends at the first recorded patch, or the calculation time when no patch is recorded. Exploitation to patch shows the interval between those two events; a published proof of concept is a public exploit, not exploitation.
Each result shows its source dates. A negative interval means the end event came first; when date-only inputs leave the order open, the result says so. Date-only inputs are approximate. The earliest evidence wins: KEV report dates and sightings are labelled as such. Records describing another vulnerability are listed but never calculated from, and a PoC alone does not establish exploitation. Missing patch evidence does not prove that no patch exists. Filters below change the event list only.
Loading timeline…
The approach is described in our paper Mapping CVEs to MITRE ATT&CK Techniques: A Curated Gold-Set Classifier and the Limits of LLM-Assisted Label Expansion.
Browse all ATT&CK techniques and the vulnerabilities related to each.
Related by attack behaviour
Vulnerabilities whose description is nearest to this one in the vector space of the CIRCL/vulnerability-attack-technique-biencoder model. This is a similarity search over the bi-encoder space (plain cosine), not a classification, and it has no measured accuracy.