GHSA-898v-775g-777c
Vulnerability from github
Published
2025-12-09 17:19
Modified
2025-12-11 15:49
Severity ?
9.4 (Critical) - CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:L/I:H/A:H
Summary
Neuron MySQLWriteTool allows arbitrary/destructive SQL when exposed to untrusted prompts (agent “footgun”)
Details

Impact

MySQLWriteTool executes arbitrary SQL provided by the caller using PDO::prepare() + execute() without semantic restrictions.

This is consistent with the name (“write tool”), but in an LLM/agent context it becomes a high-risk capability: prompt injection or indirect prompt manipulation can cause execution of destructive queries such as DROP TABLE, TRUNCATE, DELETE, ALTER, or privilege-related statements (subject to DB permissions).

Who is impacted: Deployments that expose an agent with MySQLWriteTool enabled to untrusted input and/or run the tool with a DB user that has broad privileges.

Patches

Not patched in: 2.8.11

Recommended improvements (even if keeping the tool intentionally powerful):

  • Provide a safer API that supports only constrained operations (e.g., insertRecord, updateRecord) with allowlisted tables/columns.

  • Add a policy/allowlist layer (e.g., allow only INSERT/UPDATE on selected tables; forbid DROP/TRUNCATE/ALTER/GRANT).

  • Add optional review workflow: log + require human approval for high-risk statements; or “dry-run” mode.

  • Document strongly that the tool must not be exposed to untrusted prompts without additional safeguards.

Workarounds

  • Do not enable MySQLWriteTool for public/untrusted agents.

  • Use a dedicated DB user with least privilege:

  • no DROP, no ALTER, no GRANT, no access to sensitive tables unless necessary

  • Add an application-layer policy rejecting high-risk statements (DROP, TRUNCATE, ALTER, GRANT, REVOKE, CREATE USER, etc.).

  • Implement authorization gating for tool calls (RBAC, allow tool use only for trusted operators).

Show details on source website

To clipboard 

{
  "affected": [
    {
      "database_specific": {
        "last_known_affected_version_range": "\u003c= 2.8.11"
      },
      "package": {
        "ecosystem": "Packagist",
        "name": "neuron-core/neuron-ai"
      },
      "ranges": [
        {
          "events": [
            {
              "introduced": "0"
            },
            {
              "fixed": "2.8.12"
            }
          ],
          "type": "ECOSYSTEM"
        }
      ]
    }
  ],
  "aliases": [
    "CVE-2025-67510"
  ],
  "database_specific": {
    "cwe_ids": [
      "CWE-250",
      "CWE-284"
    ],
    "github_reviewed": true,
    "github_reviewed_at": "2025-12-09T17:19:42Z",
    "nvd_published_at": "2025-12-10T23:15:48Z",
    "severity": "CRITICAL"
  },
  "details": "### Impact\n\n`MySQLWriteTool` executes arbitrary SQL provided by the caller using `PDO::prepare()` + `execute()` without semantic restrictions.  \n\nThis is consistent with the name (\u201cwrite tool\u201d), but in an LLM/agent context it becomes a high-risk capability: prompt injection or indirect prompt manipulation can cause execution of destructive queries such as `DROP TABLE`, `TRUNCATE`, `DELETE`, `ALTER`, or privilege-related statements (subject to DB permissions).\n\n\n\n**Who is impacted:** Deployments that expose an agent with `MySQLWriteTool` enabled to untrusted input and/or run the tool with a DB user that has broad privileges.\n\n### Patches\n\n**Not patched in:** 2.8.11  \n\nRecommended improvements (even if keeping the tool intentionally powerful):\n\n- Provide a safer API that supports only constrained operations (e.g., `insertRecord`, `updateRecord`) with allowlisted tables/columns.\n\n- Add a policy/allowlist layer (e.g., allow only `INSERT`/`UPDATE` on selected tables; forbid `DROP/TRUNCATE/ALTER/GRANT`).\n\n- Add optional review workflow: log + require human approval for high-risk statements; or \u201cdry-run\u201d mode.\n\n- Document strongly that the tool must not be exposed to untrusted prompts without additional safeguards.\n\n\n\n### Workarounds\n\n- Do not enable `MySQLWriteTool` for public/untrusted agents.\n\n- Use a dedicated DB user with **least privilege**:\n\n  - no `DROP`, no `ALTER`, no `GRANT`, no access to sensitive tables unless necessary\n\n- Add an application-layer policy rejecting high-risk statements (`DROP`, `TRUNCATE`, `ALTER`, `GRANT`, `REVOKE`, `CREATE USER`, etc.).\n\n- Implement authorization gating for tool calls (RBAC, allow tool use only for trusted operators).",
  "id": "GHSA-898v-775g-777c",
  "modified": "2025-12-11T15:49:17Z",
  "published": "2025-12-09T17:19:42Z",
  "references": [
    {
      "type": "WEB",
      "url": "https://github.com/neuron-core/neuron-ai/security/advisories/GHSA-898v-775g-777c"
    },
    {
      "type": "ADVISORY",
      "url": "https://nvd.nist.gov/vuln/detail/CVE-2025-67510"
    },
    {
      "type": "WEB",
      "url": "https://github.com/neuron-core/neuron-ai/commit/44bab85d92bf162898ee48d0bcef6ba0d29b59c9"
    },
    {
      "type": "PACKAGE",
      "url": "https://github.com/neuron-core/neuron-ai"
    },
    {
      "type": "WEB",
      "url": "https://github.com/neuron-core/neuron-ai/releases/tag/2.8.12"
    }
  ],
  "schema_version": "1.4.0",
  "severity": [
    {
      "score": "CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:L/I:H/A:H",
      "type": "CVSS_V3"
    }
  ],
  "summary": "Neuron MySQLWriteTool allows arbitrary/destructive SQL when exposed to untrusted prompts (agent \u201cfootgun\u201d)"
}


Log in or create an account to share your comment.




Tags
Taxonomy of the tags.




Sightings

Author Source Type Date

Nomenclature

  • Seen: The vulnerability was mentioned, discussed, or seen somewhere by the user.
  • Confirmed: The vulnerability is confirmed from an analyst perspective.
  • Published Proof of Concept: A public proof of concept is available for this vulnerability.
  • Exploited: This vulnerability was exploited and seen by the user reporting the sighting.
  • Patched: This vulnerability was successfully patched by the user reporting the sighting.
  • Not exploited: This vulnerability was not exploited or seen by the user reporting the sighting.
  • Not confirmed: The user expresses doubt about the veracity of the vulnerability.
  • Not patched: This vulnerability was not successfully patched by the user reporting the sighting.