Customer Voice Synthesizer

• Outcome: Pass
• Latest audit: May 29, 2026 at 8:21 PM UTC
• Version:
  1.0.0

Overview

This skill is a local customer-feedback summarizer that reads selected input and writes an optional report, with no active network, credential, persistence, or destructive behavior found.

Install only if you are comfortable letting the skill read the specific customer-feedback files you provide. Redact personal or sensitive customer information first, use narrow input files rather than whole workspaces, and review the generated Markdown before sharing it.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Excessive Agency: Unrestricted Tool Access, Autonomous Decision Making, Scope Creep
• MCP Least Privilege: Underdeclared Capability, Wildcard Permission, Missing Permission Declaration
• MCP Tool Poisoning: Hidden Instructions, Unicode Deception, Parameter Description Injection
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration

Findings (4)

Lp3

Medium

• Category: MCP Least Privilege
• Confidence: 84% confidence
• Finding: The skill declares no permissions, yet the content explicitly allows shell execution via `python3 "{baseDir}/scripts/run.py" --input <输入文件> --output <输出文件>` and references reading templates/specs and writing output files. This creates a capability/expectation gap: callers or platforms may treat the skill as low-risk while it can actually read local files, write artifacts, and invoke a subprocess, increasing the chance of unintended access or execution.

Description-Behavior Mismatch

High

• Confidence: 96% confidence
• Finding: The implementation materially diverges from the declared skill purpose: instead of synthesizing customer voice/JTBD inputs, it provides generic directory inspection, pattern scanning, and skill auditing capabilities. In a skill ecosystem, this kind of scope mismatch is dangerous because users may grant access to research data or repositories under false pretenses, enabling unintended file enumeration and analysis outside the expected business function.

Context-Inappropriate Capability

Medium

• Confidence: 89% confidence
• Finding: The pattern scanning feature searches arbitrary files and directories for secrets, private URLs, and risky shell snippets, which is unrelated to customer-voice synthesis. While not inherently malicious, embedding covert security-scanning behavior in a research-oriented skill increases the chance of unexpected repository inspection and exposure of sensitive content in generated reports.

Context-Inappropriate Capability

Medium

• Confidence: 88% confidence
• Finding: The skill-package compliance audit enumerates files, parses SKILL.md metadata, and reports on packaging structure, which is unrelated to the advertised JTBD/customer-voice workflow. This hidden expansion of capability undermines trust boundaries and could be used to inspect local skill repositories or project contents that users did not intend to expose to a customer research tool.

VirusTotal

65/65 vendors flagged this skill as clean.