Back to skill

Security audit

Pipeworx poetry

Security checks for vulnerabilities and agentic risk

Overview

This poetry skill is simple and purpose-aligned, but its connection instructions automatically run an unpinned npm package from the mutable latest tag.

Review before installing. Prefer a pinned and reviewed mcp-remote version or a preinstalled verified local executable, and run it with minimal filesystem, environment variable, and network access. The poetry API purpose itself appears coherent and does not request API keys.

Vulnerability Patterns
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • 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)

T08 · Insecure Dependencies

Error
Location
SKILL.md:23
Finding

Automatic Execution of an Unpinned npm Package

Content
View full analysis

Vulnerability Details

File Location: SKILL.md, lines 23–24
Vulnerability Type: Unpinned third-party dependency execution
Risk Level: High

Vulnerable Code

json
"command": "npx",
"args": ["-y", "mcp-remote@latest", "https://gateway.pipeworx.io/poetry/mcp"]

Technical Analysis

The Skill configuration invokes npx with the -y option and requests mcp-remote@latest. The -y option suppresses installation confirmation, while the mutable latest tag causes npm to retrieve and execute whichever package version is currently designated as the latest release.

Consequently, the code executed when the MCP server starts is not fixed to the version that existed during security review. If the package, its npm publishing account, or one of its transitive dependencies is compromised, a later invocation can automatically download and execute attacker-controlled JavaScript. The downloaded package executes with the operating-system identity and effective permissions of the Agent process.

No evidence establishes that the current mcp-remote package is malicious. The confirmed issue is the unsafe dependency retrieval and execution mechanism.

Attack Path

  1. An attacker compromises the mcp-remote npm package, its publisher account, the relevant supply chain, or a transitive dependency.
  2. The attacker publishes a malicious release and causes it to resolve through the latest tag.
  3. The Agent starts the configured pipeworx-poetry MCP server.
  4. npx -y retrieves the newly designated package without requiring user confirmation.
  5. npm executes the package under the Agent user's security context.
  6. The malicious package can access resources available to that process and may exfiltrate data, alter files, invoke local commands, or make unauthorized network requests.

Impact Assessment

Successful exploitation provides arbitrary code execution with the privileges of the user or service acco ...[truncated 606 chars]

Remediation
View remediation

Remediation Suggestions

  1. Replace mcp-remote@latest with an exact, reviewed version such as mcp-remote@x.y.z; do not use mutable npm tags or version ranges.
  2. Install the dependency through a committed lockfile and use a deterministic installation command such as npm ci.
  3. Verify package integrity using lockfile integrity metadata, registry provenance, and an approved package hash where supported.
  4. Review the pinned package and its complete transitive dependency tree before deployment.
  5. Disable automatic installation during normal Skill startup. Prefer invoking a previously installed and verified local executable.
  6. Run the MCP client in a sandbox or container with a read-only filesystem where practical, minimal environment variables, no unnecessary credentials, and restricted outbound network access.
  7. Use automated dependency monitoring, but require security review before updating the pinned version.
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