Security audit

照片

Security checks for vulnerabilities and agentic risk

Overview

This webcam photo skill is not clearly malicious, but it asks for broad execution and file authority while giving vague, overbroad instructions for a privacy-sensitive use case.

Review this before installing. Use it only for webcam sources you are authorized to access, and avoid capturing people, private spaces, or sensitive locations without clear consent. If installed, keep command execution and file writes constrained to explicit snapshot retrieval tasks and do not rely on the unrelated security-analysis or CI/CD claims.

SkillSpector

Vulnerability Patterns

Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
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 (8)

Intent-Code Divergence

Medium

Confidence: 92% confidence
Finding: The skill markets paid-only static analysis, vulnerability detection, report generation, and CI/CD features that are unrelated to its stated webcam-snapshot purpose. This kind of capability inflation can mislead users and agents into granting broader trust, invoking the skill in inappropriate contexts, or expecting security functions that do not actually exist.

Intent-Code Divergence

Low

Confidence: 85% confidence
Finding: The typical scenarios section mixes generic scraping and network configuration tasks into a webcam-photo skill, broadening the apparent scope beyond what is justified. Ambiguous scope increases the chance that an agent will route unrelated tasks to a skill that has read/write/exec access, creating unnecessary exposure.

Context-Inappropriate Capability

Medium

Confidence: 95% confidence
Finding: The skill declares exec capability even though the documentation does not justify why system command execution is necessary for listing and retrieving webcam snapshots. In a skill with vague scope and permissive triggers, unnecessary command execution materially raises the risk of shell misuse, unintended local actions, or abuse through future prompt expansion.

Intent-Code Divergence

Medium

Confidence: 90% confidence
Finding: The core feature list advertises broad file handling, API integration, command execution, and information retrieval capabilities beyond the webcam-snapshot intent. This overbroad framing can cause over-privileging and unsafe task routing, especially because the skill already exposes powerful tools like read, write, and exec.

Vague Triggers

Medium

Confidence: 88% confidence
Finding: The activation guidance says to use the skill whenever the user needs photo-related functionality, which is far broader than the documented webcam snapshot purpose. Overly broad triggers increase accidental invocation, exposing read/write/exec tools in contexts where they are unnecessary and potentially unsafe.

Vague Triggers

Medium

Confidence: 82% confidence
Finding: The usage guidance contains an incomplete and ambiguous trigger phrase, making it unclear when the skill should activate. Ambiguity in activation logic is risky for agentic systems because it can lead to misrouting and inappropriate use of privileged tools.

Missing User Warnings

High

Confidence: 97% confidence
Finding: A skill for retrieving webcam snapshots lacks a prominent upfront warning about privacy, consent, surveillance, and legality constraints. In this context, the omission is significant because webcam imagery can involve personal data and real-world monitoring, making misuse more likely and more harmful than for ordinary web content.

Natural-Language Policy Violations

Medium

Confidence: 80% confidence
Finding: The description states that the skill supports Chinese interaction and frames usage around Chinese-language interaction without clear user opt-in. While not a direct code-execution issue, forced language behavior can reduce user comprehension of privacy/legal warnings and increase operational mistakes in a sensitive webcam-monitoring context.

Static analysis

No suspicious patterns detected.