Back to skill

Security audit

MnemoPay

Security checks for vulnerabilities and agentic risk

Overview

The skill is coherent, but it directs users to run a money and identity MCP service from unpinned packages, which needs review before installation.

Review before installing. Use only the minimum needed tool group, avoid admin or operator groups for ordinary agents, require explicit user approval for payments and high-impact actions, and prefer a pinned, reviewed SDK version or fixed local executable rather than unattended npx -y package resolution.

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:12
Finding

Unpinned Third-Party Packages Are Downloaded and Executed

Content
View full analysis

Vulnerability Details

File Location: SKILL.md, lines 12–17
Vulnerability Type: Unverified and unpinned dependency installation
Risk Level: High

bash
npm install @mnemopay/sdk
claude mcp add -s user mnemopay -- npx -y @mnemopay/sdk mcp

Python:

bash
pip install mnemopay

Technical Analysis

The installation instructions retrieve npm and Python packages without pinning an exact reviewed version or verifying package integrity. In particular, npx -y @mnemopay/sdk mcp automatically approves package retrieval and executes the version resolved by the package registry at invocation time.

The Claude command also creates a user-scoped MCP registration that refers to this mutable npx package resolution. Consequently, a package version reviewed at one point in time may differ from the version executed during a later MCP launch. The project does not provide a lockfile, checksum, signature-verification procedure, or fixed local executable that would bind execution to a known artifact.

This is a supply-chain weakness rather than evidence that the current MnemoPay packages are malicious. Exploitation would require compromise of the relevant package, publisher account, registry resolution, or transitive dependency.

Attack Path

  1. An attacker compromises the npm or Python publisher account, package distribution path, or a dependency used by the package.
  2. The attacker publishes a malicious release that remains compatible with the unpinned package name.
  3. A user follows the documented installation command, or Claude launches the user-scoped MCP registration through npx.
  4. The package manager resolves and downloads the attacker-controlled release without requiring an exact version or expected integrity value.
  5. Installation hooks or package runtime code execute with the privileges of the invoking user.
  6. The malicious code can access resources available to that user and may impersona ...[truncated 825 chars]
Remediation
View remediation

Remediation Suggestions

  1. Pin npm and Python dependencies to exact, security-reviewed versions rather than package names or floating ranges.
  2. Generate and retain lockfiles with integrity metadata, such as package-lock.json for npm and a hash-locked requirements file for Python.
  3. Verify downloaded artifacts through registry integrity hashes, trusted signatures, or independently published checksums.
  4. Avoid registering an MCP command that uses unattended npx -y resolution on every launch. Install a reviewed version once and register its fixed local executable path.
  5. Disable install scripts where feasible, audit transitive dependencies, and scan package releases before deployment.
  6. Run the MCP service under a dedicated least-privileged account or sandbox with restricted filesystem, network, secret, and credential access.
  7. Continue exposing only the minimum necessary MCP tool groups, and require explicit authorization for administrative or operator capabilities.
  8. Document how users can remove or update the user-scoped MCP registration safely when a release is revoked or compromised.
Vulnerability Patterns
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Findings (4)

Rp1

Medium
Category
MCP Rug Pull
Confidence
93% confidence
Finding

The skill instructs users to launch an MCP server via npx -y @mnemopay/sdk without pinning an exact package version. This creates a supply-chain risk: a future compromised, malicious, or breaking release could be pulled and executed automatically at install/run time. Because this skill is explicitly for money handling, approvals, identity, and consequential actions, compromise of the fetched package would be especially dangerous.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
93% confidence
Finding

This command references @mnemopay/sdk without an exact version, allowing execution of whatever package version is current when the command is run. If the npm package or publisher account were compromised, an attacker could deliver malicious MCP server code to agents with access to governance, identity, or skill-related capabilities. The financial and administrative context increases the blast radius.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
93% confidence
Finding

Running npx -y @mnemopay/sdk without version pinning exposes consumers to unreviewed upstream changes and potential supply-chain compromise. In this line, the enabled tool groups include governance and identity functions, so a malicious package update could misuse privileged workflows, approvals, identity data, or organizational controls.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
94% confidence
Finding

This unpinned npx invocation is especially risky because it enables the agent_os tool group, which likely exposes durable browser, code, or computer-use jobs. If an attacker controlled a future package version, they could gain execution in a highly sensitive automation environment and potentially trigger consequential actions or access governed memory and payment-related operations.

Content

No source excerpt is available for this finding.