Back to skill

Security audit

Shipyard design

Security checks for vulnerabilities and agentic risk

Overview

ClawScan marked this skill suspicious because it reports an instruction that could lead to commits and PRs without a clear approval checkpoint, and unpinned package commands that may run changing releases; the scan was incomplete.

Review the skill’s approval steps before installing, especially the reported commit and PR instructions. Treat the unpinned package commands as a supply-chain risk, since consent to install does not pin the version that runs. Skillspector coverage was partial: it scanned 12 of 13 components, with one file partly inspected, so the evidence is incomplete. The verdict is suspicious, not a claim that the author acted maliciously.

Vulnerability Patterns
  • 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
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Vulnerability Patterns
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • 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
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Findings (6)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
88% confidence
Finding

The supplied code implements sound effects rather than visual styling. Although sounds can complement interface design, audio playback is a distinct user-facing capability not disclosed by the appearance-focused description. The preference lookup and audio caching are supporting implementation details, not separate mismatches.

Content

No source excerpt is available for this finding.

The tidy-themes workflow instructs the agent to commit and open a PR without preserving the explicit ask-first requirement.

Medium
Category
Excessive Agency
Confidence
86% confidence
Finding

The tidy-themes workflow directly instructs the agent to commit and open a PR, conflicting with the global requirement to ask before committing or pushing. Although that global rule offers some protection, the specific workflow creates ambiguity that could cause unauthorized repository changes and remote publication of project changes.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'npx impeccable'.

Medium
Category
MCP Rug Pull
Confidence
89% confidence
Finding

This is a CLI installer rather than an MCP server, but the unpinned npx command can download and execute a changing package release with the developer's permissions. Asking before installation reduces unauthorized execution risk, but does not ensure the approved package version is the one executed; the documented installation of edit hooks also makes package trust consequential.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'npx impeccable'.

Medium
Category
MCP Rug Pull
Confidence
89% confidence
Finding

The unpinned npx impeccable update command can execute a newly resolved package release and refresh installed agent tooling. Although this is not an MCP server and installation is consent-gated, neither that consent nor the README verification establishes integrity of future releases.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'npx ui-ux-pro-max-cli'.

Medium
Category
MCP Rug Pull
Confidence
89% confidence
Finding

The fallback npx command executes an unpinned third-party CLI, and the adjacent global npm installation is also unpinned. These are not MCP server declarations, but a compromised future package release could execute with the user's filesystem access; the optional, ask-first workflow reduces exposure without addressing release integrity.

Content

No source excerpt is available for this finding.

Generic activation phrases such as "design" and "3 directions" can trigger the skill outside app-appearance requests.

Low
Category
Trigger Abuse
Confidence
88% confidence
Finding

Although the description states an app-appearance purpose, its explicit trigger list includes standalone phrases that also occur in architecture, graphic design, and ordinary directional requests. It does not explicitly require those generic phrases to refer to an app's visual appearance or exclude unrelated uses.

Content

No source excerpt is available for this finding.