Back to skill

Security audit

Pipeworx core-research

Security checks for vulnerabilities and agentic risk

Overview

This is a small CORE research connector with no bundled code or persistence, but its research queries are routed through a pipeworx.io MCP gateway.

Install only if you are comfortable sending research queries to the configured pipeworx.io MCP gateway. Avoid submitting confidential research topics, proprietary prompts, credentials, or personal data unless the gateway operator and its relationship to CORE are verified.

Vulnerability Patterns
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
Findings (1)

T09 · Insecure Skill Coding Practices

Warning
Location
SKILL.md:14
Finding

Research Requests Routed Through an Unverified Third-Party MCP Endpoint

Content
View full analysis

Vulnerability Details

File Location: SKILL.md, lines 14-21
Vulnerability Type: Untrusted external service configuration
Risk Level: Medium

The skill describes integration with the CORE API but configures its MCP server through the unrelated domain gateway.pipeworx.io rather than a documented CORE-operated endpoint.

json
{
  "mcpServers": {
    "core-research": {
      "url": "https://gateway.pipeworx.io/core-research/mcp"
    }
  }
}

Technical Analysis

The configured MCP endpoint forms a privileged trust boundary: the agent sends research requests to it and consumes responses as tool output. Although HTTPS protects data in transit, it does not establish that pipeworx.io is operated or authorized by CORE.

If this intermediary is malicious, compromised, transferred to another operator, or incorrectly configured, it can observe submitted queries and return attacker-controlled MCP responses. The reviewed file does not document endpoint ownership, response validation, transmitted-data restrictions, or a direct trusted CORE endpoint.

No evidence in the reviewed artifact establishes that the endpoint currently executes code or is malicious. The confirmed issue is the undocumented third-party trust and disclosure exposure created by the configuration.

Attack Path

  1. A user installs the skill under the assumption that it communicates with CORE.
  2. The agent connects to https://gateway.pipeworx.io/core-research/mcp.
  3. Research queries and associated MCP request metadata are disclosed to the third-party service.
  4. A malicious or compromised service records those requests or returns crafted results.
  5. The agent or user relies on manipulated research metadata, abstracts, or full-text content as if it came from the advertised CORE integration.

Impact Assessment

The endpoint operator can potentially access all information submitted through the research tool, including sensi ...[truncated 518 chars]

Remediation
View remediation

Remediation Suggestions

  • Use an endpoint operated and documented by CORE whenever an official MCP or API endpoint is available.
  • If the intermediary is required, clearly disclose its operator, purpose, data-handling policy, retention policy, and relationship with CORE.
  • Verify endpoint ownership and pin the configuration to an organization-controlled allowlist.
  • Restrict requests to non-sensitive research terms and prevent credentials, personal data, proprietary prompts, or unrelated conversation context from being forwarded.
  • Treat all MCP responses as untrusted data and ensure returned text cannot be interpreted as agent instructions.
  • Validate response schemas, constrain response sizes, and sanitize rendered content.
  • Add monitoring and an incident-response mechanism for endpoint ownership, certificate, or behavior changes.
  • Conduct a security and privacy review of the service provider before distributing the skill.
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep