Security audit

页面复刻分析器

Security checks for vulnerabilities and agentic risk

Overview

This skill coherently analyzes user-provided screenshots for page cloning, but users should avoid submitting screenshots that contain secrets or private data.

Install only if you are comfortable with the agent transcribing all visible text in screenshots you provide. Before using it, crop or redact passwords, tokens, customer data, internal URLs, financial records, or other confidential information unless you explicitly want those values included in the page spec.

SkillSpector

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep

Findings (5)

Missing User Warnings

High

Confidence: 97% confidence
Finding: The skill explicitly requires verbatim extraction of all visible screenshot text and data, which can include PII, credentials, internal URLs, ticket numbers, financial data, or other confidential content. Because it provides no warning, minimization, redaction, or consent step, it increases the likelihood that sensitive data will be reproduced and retained in downstream outputs.

Missing User Warnings

High

Confidence: 96% confidence
Finding: The output template mandates preserving all visible text '一字不改', which operationally instructs the agent to reproduce any sensitive content embedded in screenshots. In a page-cloning context, this is particularly risky because screenshots of enterprise systems often contain real production data, making disclosure more likely and more damaging.

Ssd 3

Medium

Confidence: 95% confidence
Finding: The skill instructs the model to extract and preserve all visible on-screen data from user screenshots, which materially increases the chance of exposing personal or confidential information. Although the data originates from the user, the unsafe part is the unconditional, exhaustive reproduction without classification, filtering, or minimization.

Ssd 3

Medium

Confidence: 95% confidence
Finding: A mandatory complete data inventory that retains all text exactly as shown encourages the model to surface potentially sensitive screenshot contents wholesale into the final response. That behavior can amplify exposure by moving data from an image into searchable, copyable text form.

Ssd 3

Medium

Confidence: 94% confidence
Finding: The core principle requiring complete verbatim retention of all screenshot text reinforces broad disclosure and overrides safer data-minimization behavior. In this skill's context of 1:1 page cloning, that makes leakage more dangerous because realistic screenshots frequently contain authentic operational data rather than placeholder content.

Static analysis

No suspicious patterns detected.