GHSA-647H-P824-99W7

Vulnerability from github – Published: 2026-03-25 17:23 – Updated: 2026-03-25 17:23
VLAI
Summary
@grackle-ai/mcp has a workspace authorization bypass in its knowledge_search MCP tool
Details

Impact

The knowledge_search and knowledge_get_node MCP tools are included in SCOPED_TOOLS (visible to scoped agents) but their handlers do not receive authContext and do not enforce workspace scoping. A scoped agent in Workspace A can supply an arbitrary workspaceId parameter to search or retrieve knowledge graph nodes from Workspace B, bypassing workspace isolation boundaries.

This is a cross-workspace data leakage vulnerability affecting any deployment where multiple workspaces contain sensitive knowledge graph data and scoped agents are used.

Affected code: - packages/mcp/src/tools/knowledge.ts:146-169 (knowledge_search handler) - packages/mcp/src/tools/knowledge.ts:244-283 (knowledge_get_node handler) - packages/mcp/src/tool-scoping.ts:11 (both tools listed in SCOPED_TOOLS)

Contrast with correct implementation: knowledge_create_node (same file, lines 334-357) properly receives authContext and overrides the user-supplied workspaceId for scoped callers.

Design Note

Cross-workspace knowledge sharing is a legitimate future feature — agents working across different repos may need to collaborate and share knowledge. However, this access should be opt-in with explicit grants, not an implicit bypass. The immediate fix locks scoped agents to their own workspace. A future design could introduce: - Workspace-level "share knowledge with" settings - A cross_workspace scope on scoped tokens - Explicit workspaceIds (plural) in the auth context

Patches

Fix: Add authContext parameter to knowledge_search and knowledge_get_node handlers and enforce workspace scoping, matching the pattern in knowledge_create_node:

const resolvedWorkspaceId =
  authContext?.type === "scoped"
    ? authContext.workspaceId ?? ""
    : workspaceId ?? "";

When cross-workspace collaboration is designed, this check can be relaxed intentionally with proper access controls.

Workarounds

Do not use scoped agent tokens in multi-workspace deployments until patched. Alternatively, remove knowledge_search and knowledge_get_node from the SCOPED_TOOLS set in tool-scoping.ts.

References

  • CWE-284: Improper Access Control
  • File: packages/mcp/src/tools/knowledge.ts
  • File: packages/mcp/src/tool-scoping.ts
Show details on source website

{
  "affected": [
    {
      "database_specific": {
        "last_known_affected_version_range": "\u003c= 0.70.1"
      },
      "package": {
        "ecosystem": "npm",
        "name": "@grackle-ai/mcp"
      },
      "ranges": [
        {
          "events": [
            {
              "introduced": "0"
            },
            {
              "fixed": "0.70.2"
            }
          ],
          "type": "ECOSYSTEM"
        }
      ]
    }
  ],
  "aliases": [],
  "database_specific": {
    "cwe_ids": [
      "CWE-284"
    ],
    "github_reviewed": true,
    "github_reviewed_at": "2026-03-25T17:23:11Z",
    "nvd_published_at": null,
    "severity": "HIGH"
  },
  "details": "### Impact\n\nThe `knowledge_search` and `knowledge_get_node` MCP tools are included in `SCOPED_TOOLS` (visible to scoped agents) but their handlers do not receive `authContext` and do not enforce workspace scoping. A scoped agent in Workspace A can supply an arbitrary `workspaceId` parameter to search or retrieve knowledge graph nodes from Workspace B, bypassing workspace isolation boundaries.\n\nThis is a **cross-workspace data leakage** vulnerability affecting any deployment where multiple workspaces contain sensitive knowledge graph data and scoped agents are used.\n\n**Affected code:**\n- `packages/mcp/src/tools/knowledge.ts:146-169` (knowledge_search handler)\n- `packages/mcp/src/tools/knowledge.ts:244-283` (knowledge_get_node handler)\n- `packages/mcp/src/tool-scoping.ts:11` (both tools listed in SCOPED_TOOLS)\n\n**Contrast with correct implementation:** `knowledge_create_node` (same file, lines 334-357) properly receives `authContext` and overrides the user-supplied `workspaceId` for scoped callers.\n\n### Design Note\n\nCross-workspace knowledge sharing is a legitimate future feature \u2014 agents working across different repos may need to collaborate and share knowledge. However, this access should be **opt-in with explicit grants**, not an implicit bypass. The immediate fix locks scoped agents to their own workspace. A future design could introduce:\n- Workspace-level \"share knowledge with\" settings\n- A `cross_workspace` scope on scoped tokens\n- Explicit `workspaceIds` (plural) in the auth context\n\n### Patches\n\n**Fix:** Add `authContext` parameter to `knowledge_search` and `knowledge_get_node` handlers and enforce workspace scoping, matching the pattern in `knowledge_create_node`:\n```typescript\nconst resolvedWorkspaceId =\n  authContext?.type === \"scoped\"\n    ? authContext.workspaceId ?? \"\"\n    : workspaceId ?? \"\";\n```\n\nWhen cross-workspace collaboration is designed, this check can be relaxed intentionally with proper access controls.\n\n### Workarounds\n\nDo not use scoped agent tokens in multi-workspace deployments until patched. Alternatively, remove `knowledge_search` and `knowledge_get_node` from the `SCOPED_TOOLS` set in `tool-scoping.ts`.\n\n### References\n\n- CWE-284: Improper Access Control\n- File: `packages/mcp/src/tools/knowledge.ts`\n- File: `packages/mcp/src/tool-scoping.ts`",
  "id": "GHSA-647h-p824-99w7",
  "modified": "2026-03-25T17:23:11Z",
  "published": "2026-03-25T17:23:11Z",
  "references": [
    {
      "type": "WEB",
      "url": "https://github.com/nick-pape/grackle/security/advisories/GHSA-647h-p824-99w7"
    },
    {
      "type": "PACKAGE",
      "url": "https://github.com/nick-pape/grackle"
    }
  ],
  "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:L/SI:L/SA:N",
      "type": "CVSS_V4"
    }
  ],
  "summary": "@grackle-ai/mcp has a workspace authorization bypass in its knowledge_search MCP tool"
}



Log in or create an account to share your comment.




Tags
Taxonomy of the tags.


Loading…

Loading…

Loading…

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.

Loading…

Loading…

Loading…

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.


Loading…