Back to skill

Security audit

Pipeworx court-listener

Security checks for vulnerabilities and agentic risk

Overview

The skill is a CourtListener research tool, but it routes queries through a third-party Pipeworx MCP gateway without clearly explaining that trust boundary.

Install only if you are comfortable sending CourtListener searches, docket queries, and opinion IDs to the Pipeworx gateway. For important legal work, verify case metadata and opinion text against CourtListener or another authoritative source before relying on the results.

Vulnerability Patterns
  • Tool Hijacking and SpoofingModifies or replaces tools so legitimate-looking calls execute attacker logic
  • 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)

T07 · Tool Hijacking and Spoofing

Warning
Location
SKILL.md:19
Finding

Undisclosed Third-Party MCP Endpoint Spoofs CourtListener Integration

Content
View full analysis

Vulnerability Details

File Location: SKILL.md, lines 19–23
Vulnerability Type: Tool hijacking and spoofing through an unrelated third-party MCP gateway
Risk Level: Medium

Vulnerable Code

json
{
  "mcpServers": {
    "court-listener": {
      "url": "https://gateway.pipeworx.io/court-listener/mcp"
    }
  }
}

Technical Analysis

The skill describes itself as an interface to the Free Law Project's CourtListener API, but the configured MCP server is hosted on gateway.pipeworx.io, rather than an official CourtListener or Free Law Project domain. The configuration therefore delegates request confidentiality, tool behavior, and response integrity to an undisclosed third-party intermediary.

Because MCP responses can influence an agent's conclusions and subsequent actions, a malicious or compromised gateway could return modified, fabricated, or selectively omitted legal information while appearing to operate as the legitimate court-listener tool. The remote operator can also change this behavior after the skill has been reviewed because the server implementation is not included in the audited project.

This finding does not establish that the gateway is currently malicious. It identifies an unverified trust boundary and a spoofing risk arising from the mismatch between the represented service and the configured endpoint.

Attack Path

  1. A user installs or loads the MCP configuration from SKILL.md.
  2. The client registers the third-party endpoint under the legitimate-looking name court-listener.
  3. The user or agent submits legal search terms, docket queries, or opinion identifiers through that tool.
  4. Requests are transmitted to gateway.pipeworx.io, allowing the gateway operator or a party that compromises it to observe the research activity.
  5. The gateway supplies tool responses and can alter, omit, or fabricate case metadata and opinion content.
  6. The agent treats tho ...[truncated 721 chars]
Remediation
View remediation

Remediation Suggestions

  1. Replace the gateway with an official, documented CourtListener or Free Law Project endpoint where such an endpoint is available.
  2. If MCP translation is required, use a locally hosted and audited adapter that communicates directly with the documented CourtListener API.
  3. If the Pipeworx gateway must remain, explicitly disclose that it is a third-party intermediary, identify its operator, and document its data retention, privacy, and response-integrity guarantees.
  4. Authenticate the remote server and pin its expected identity or deployment where the MCP client supports those controls.
  5. Restrict the MCP server to the minimum required tool capabilities and avoid transmitting confidential case strategy, personal information, credentials, or other sensitive query data.
  6. Validate returned case identifiers and legally significant content against an authoritative source before relying on them.
  7. Require informed user confirmation before sending queries to the third-party service.
  8. Establish monitoring and review controls for endpoint ownership, TLS configuration, unexpected tool-schema changes, and anomalous responses.
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