/

CVE-2026-55609

Severity:

HIGH RISK

(7.1)

Path Traversal (CWE-73) in Sublinear Solver MCP Tools

CVE-2026-55609

Deferred

Vulnerability type

Category:

Path Traversal

CWE:

CWE-73

Impact:

Arbitrary File Read and Write

Severity:

HIGH RISK

(7.1)

CVSS version:

3.1

Catch this CVE while code is being written

Detect vulnerable code directly in pull requests, so CVEs are fixed before merge.

Summary

A path traversal vulnerability in the MCP file-management tools of sublinear-time-solver and consciousness-explorer allows attacker-controlled file paths to reach filesystem operations without adequate destination restrictions.

An attacker who can invoke the affected MCP tools may read, create, modify, or overwrite files accessible to the server process, potentially causing data exposure, integrity loss, or service disruption.

Affected context

The vulnerability occurs in deployments that expose the affected MCP tools for saving and loading state or vector data. The filepath and file_path parameters can reach filesystem operations without sufficient path restriction or traversal rejection, and the risk depends on the permissions and filesystem access available to the server process.

Exploit preconditions

  • Attack surface: MCP server tools that perform state or vector file operations
  • Required access: Permission to invoke the affected MCP tools, either directly or through an exposed integration
  • Trigger point: Processing of attacker-controlled filepath or file_path values
  • Why exploitation is possible: File destinations are not adequately confined to an approved directory, allowing path traversal to reach other process-accessible files

Why this happens

The underlying bug is improper restriction of user-controlled file paths. The affected tools pass attacker-supplied path values into file read and write operations without reliably enforcing a trusted base directory or rejecting traversal components.

This behavior maps to CWE-73: External Control of File Name or Path, where external input influences filesystem targets without sufficient canonicalization, validation, or containment checks.

Potential impact

  • Sensitive files may be read and disclosed
  • Application or configuration files may be modified or overwritten
  • The service may be disrupted through destructive or invalid file operations
  • Credentials, tokens, or deployment data could be exposed
  • Modified files could enable further compromise depending on their use

General mitigation guidance

Mitigation generally focuses on reducing trust in file path parameters and enforcing filesystem containment.

Common defensive patterns include:

  • Resolving paths against a fixed, approved directory and verifying that the result remains within that directory
  • Rejecting traversal components, absolute paths, unexpected encodings, and platform-specific alternate path forms
  • Prefering opaque file identifiers or server-generated filenames over caller-supplied paths
  • Separating read and write permissions and allowing only the minimum required file operations
  • Running the MCP server with a dedicated low-privilege account and a restricted filesystem view
  • Applying allowlists for permitted files, extensions, and operation types
  • Treating state and vector storage locations as controlled resources rather than arbitrary filesystem targets

How teams detect issues like this

Organizations can look for this class of issue through:

  • Static analysis rules that identify external values flowing into filesystem APIs
  • Code review checks for path parameters passed to file reads, writes, renames, or deletion functions
  • Tests covering absolute paths, traversal sequences, encoded traversal, symlink behavior, and platform-specific path formats
  • Runtime audit logging for MCP tool calls, resolved destinations, and file-access failures
  • Policy enforcement that limits service processes to approved directories and denies access to sensitive paths
  • Automated security review tools that flag missing canonicalization and directory-containment checks

Sources

SourceLink
github.comgithub.com/BruceJqs/public_exp/issues/32
github.comgithub.com/ruvnet/sublinear-time-solver/commit/a701296e363192be863e79d788fa268095e3d229
github.comgithub.com/ruvnet/sublinear-time-solver/commit/ea9a212b69e4449ec443fe088a7aec7546f70b4a
github.comgithub.com/ruvnet/sublinear-time-solver/issues/19
github.comgithub.com/ruvnet/sublinear-time-solver/pull/20
github.comgithub.com/ruvnet/sublinear-time-solver/releases/tag/v1.6.0
github.comgithub.com/ruvnet/sublinear-time-solver/security/advisories/GHSA-xc9g-j69q-37xw

Frequently asked

questions

It can be remotely exploitable when the MCP server or an integration exposes the affected tools to an attacker. If tool invocation is limited to trusted local users, exploitation scope is correspondingly narrower.

The attacker needs the ability to invoke the vulnerable MCP tools and provide a controlled path parameter. The resulting impact is constrained by the permissions of the server process.

Constrain every resolved path to a fixed, approved directory, reject traversal and absolute-path forms, and use server-generated identifiers instead of accepting arbitrary filesystem paths.

Combine path-containment validation with least-privilege execution, restricted filesystem permissions, allowlisted storage locations, and automated tests for traversal and symlink-related edge cases.

Find out what's already
exploitable in your codebase.