Insight Radar (洞察雷达)

• Outcome: Pass
• Latest audit: May 29, 2026 · 8:29 PM UTC
• Version: 1.0.6

Overview

This is a disclosed personalized news-briefing skill that searches the web and stores scoped local news notes, with privacy considerations around USER.md personalization.

Before installing, review the two dependency skills, keep USER.md limited to information you are comfortable seeing in briefings or local logs, and periodically inspect or prune memory/news-log and memory/knowledge-base. Treat any separate HEARTBEAT, cron, or Feishu delivery setup as a separate integration to review because this skill itself declares no direct external output or credentials.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access
• Supply Chain: Unpinned Dependencies, External Script Fetching, Obfuscated Code
• Excessive Agency: Unrestricted Tool Access, Autonomous Decision Making, Scope Creep

Findings (6)

Missing User Warnings

Medium

• Confidence: 93% confidence
• Finding: The README explicitly describes persistent storage of analyzed news to local knowledge-base files and optional transmission to Feishu Bitable, but it does not clearly warn users about retention, overwrite behavior, or external data sharing. In an agent skill, undocumented persistence and outbound syncing can expose sensitive context from USER.md or derived analysis to locations the user did not expect, increasing privacy and data-governance risk.

Missing User Warnings

Medium

• Confidence: 94% confidence
• Finding: The skill persistently writes analyzed content into local knowledge-base files and daily logs, but does not present a clear user-facing warning or consent boundary before doing so. This creates a privacy and data-retention risk because users may request a briefing without realizing their content and derived inferences will be stored long-term.

Missing User Warnings

Medium

• Confidence: 96% confidence
• Finding: The skill instructs the agent to read USER.md for personalization, but does not clearly warn the user that personal profile data will be accessed and used to tailor output. Even if the file is local, silently consuming personal context can violate user expectations and expose sensitive details in generated briefings or stored artifacts.

Natural-Language Policy Violations

Medium

• Confidence: 90% confidence
• Finding: The example output hard-codes Chinese-language presentation and address style without any indication that this is conditional on user preference, locale, or explicit opt-in. In an agent skill, examples often shape downstream behavior, so this can cause the agent to override user language expectations, reduce accessibility, and create inappropriate or exclusionary responses.

Ssd 3

Medium

• Confidence: 97% confidence
• Finding: The skill requires reading personal user context and weaving it into generated analysis, while also writing logs and knowledge artifacts to disk. In this context, that combination is more dangerous because sensitive personal details can be transformed into derived insights, echoed in outputs, and inadvertently retained in persistent files beyond the original request.

Ssd 3

Medium

• Confidence: 98% confidence
• Finding: This section explicitly instructs the agent to extract profession, interests, challenges, and future plans from USER.md and reuse them in user-facing output. Because the skill also maintains persistent news logs and knowledge-base files, the context increases risk of privacy leakage, profiling, and accidental retention or disclosure of sensitive personal information through generated content or downstream storage.

VirusTotal

65/65 vendors flagged this skill as clean.

Static analysis

No suspicious patterns detected.