Back to skill

Security audit

Mockplus Reader

Security checks for vulnerabilities and agentic risk

Overview

The skill mostly reads MockPlus design links, but it also includes unrelated local project run instructions that can install and execute npm tooling outside the stated reader purpose.

Review this skill before installing. It is reasonable for a MockPlus reader to open user-provided MockPlus links, but avoid following the included npm/global-install project commands unless you separately trust and inspect the lutixia project, its package files, and the npm packages involved. Use a disposable environment if you run those commands.

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

Warning
Location
SKILL.md:80
Finding

Unpinned Global Dependency Installation and Execution of Unaudited Project Dependencies

Content
View full analysis

Vulnerability Details

File Location: SKILL.md, lines 80-86
Vulnerability Type: Insecure third-party dependency installation
Risk Level: Medium

Vulnerable Code

bash
**Run instructions:**
```bash
cd lutixia
npm install -g @dcloudio/uni-cli
npm install
npm run dev:%PLATFORM%
text

### Technical Analysis

The Skill instructs the user or agent to install `@dcloudio/uni-cli` globally without specifying a trusted version or integrity constraint. This causes npm to resolve the package version available from the configured registry at execution time, meaning the installed code can differ from what was present when the Skill was audited.

The subsequent `npm install` command operates on a separate `lutixia` project that is not included in the audited artifact. Its manifest, lockfile, dependency sources, version constraints, and npm lifecycle scripts therefore cannot be verified. npm installation can execute package lifecycle hooks such as `preinstall`, `install`, and `postinstall` with the invoking user's privileges.

These installation and execution steps are unrelated to the Skill's primary stated function of opening and analyzing MockPlus design pages. The global installation also modifies the user's environment beyond the requirements of that function.

### Attack Path

1. A user or agent loads the MockPlus reader Skill and follows its project run instructions.
2. `npm install -g @dcloudio/uni-cli` retrieves an unpinned package version from the configured npm registry and installs it globally.
3. If the package, one of its transitive dependencies, or the configured registry is compromised, attacker-controlled lifecycle code can execute during installation.
4. The agent then runs `npm install` inside the external `lutixia` directory.
5. Dependencies or lifecycle scripts defined by that unaudited project execute with the permissions of the invoking account.
6. The resulting project command is e
...[truncated 859 chars]
Remediation
View remediation

Remediation Suggestions

  1. Remove the implemented-project and npm execution instructions from this reader Skill because they are not necessary for opening or analyzing MockPlus pages.
  2. If the project instructions are required, include the relevant package.json and lockfile within the review scope.
  3. Pin all direct and transitive dependencies to reviewed versions using a committed lockfile and use npm ci rather than an unconstrained npm install.
  4. Avoid global installation. Install the CLI as a pinned project-local development dependency and invoke it through an explicit package script or npx --no-install.
  5. Verify package provenance and registry configuration, and use npm integrity metadata and dependency auditing.
  6. Disable lifecycle scripts during installation with --ignore-scripts when they are not required. Review and explicitly allow any necessary lifecycle scripts before execution.
  7. Run dependency installation and development commands in a restricted, disposable environment without production credentials or sensitive filesystem access.
  8. Do not instruct the agent to execute commands in an external hard-coded workspace unless that workspace and its dependency manifests have been separately audited.
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (5)

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The activation description is broad enough to trigger on generic requests about analyzing prototype drafts, which can cause the agent to open external links or process third-party content more often than the user intended. In a browser-capable skill, overbroad activation increases exposure to untrusted remote content and can cause unnecessary data handling or context switching.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
85% confidence
Finding

The document claims the skill only reads and analyzes MockPlus pages, but later content indicates an additional purpose of producing and running an application project. This intent mismatch is dangerous because security controls and user expectations rely on accurate scope declarations; hidden or undocumented side effects can lead to over-privileged use and accidental execution of local development workflows.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The skill is presented as a read-only MockPlus analysis tool, but the latter section instructs users to operate a generated uni-app/Vue3 project and exposes a concrete workspace path. This creates capability drift: an analyst or agent may treat the skill as passive while it implicitly expands into code-generation and local-environment interaction, increasing the chance of unintended command execution or misuse of local resources.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
77% confidence
Finding

The skill instructs the browser to open external MockPlus links and snapshot their contents without warning that this accesses third-party content that may require authentication, contain sensitive prototype data, or present untrusted material. While common for link-reader skills, the lack of explicit warning or consent language weakens safe handling of external resources.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
80% confidence
Finding

The documentation reveals a specific local filesystem path tied to a user's environment. Even if not directly exploitable on its own, exposing host-specific paths leaks environment details and can aid fingerprinting, targeted social engineering, or accidental reuse of unsafe, non-portable instructions.

Content

No source excerpt is available for this finding.