Security audit

AI科技洞察罗盘

Security checks across malware telemetry and agentic risk

Overview

The skill is a coherent AI-news article workflow, but it can write files and publish generated content externally without clear confirmation safeguards.

Review before installing. Use it only when you explicitly want an AI tech article workflow that may save under /root/articles and publish or archive generated content externally. Require a draft-only default or a manual confirmation step before any POST to the article API or Feishu document creation, and change the output path to a workspace-controlled location if possible.

SkillSpector

By NVIDIA

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (8)

Missing User Warnings

Medium

Confidence: 89% confidence
Finding: The README explicitly directs the workflow to save generated output to a fixed absolute path under /root without any notice, confirmation, or discussion of filesystem side effects. In an agent skill context, undocumented writes can surprise operators, overwrite existing content, or normalize running with elevated privileges, which increases risk even if the content being written is only an article.

Missing User Warnings

Medium

Confidence: 93% confidence
Finding: The skill advertises POSTing generated content to a public-account API but does not warn that article text and possibly associated metadata will be transmitted to an external service. In an agent workflow, silent outbound publication can leak sensitive prompts, unpublished analysis, credentials-adjacent metadata, or cause unintended public distribution from a connected account.

Missing User Warnings

Medium

Confidence: 90% confidence
Finding: The README describes creating a Feishu document archive without disclosing that generated content will be copied into a third-party collaboration platform, which creates account-side effects and external data exposure. In this skill's context, the generated article may include proprietary technical analysis or source-derived content, so undocumented synchronization increases confidentiality and governance risk.

Vague Triggers

Medium

Confidence: 94% confidence
Finding: The trigger phrases are broad enough to match ordinary requests for AI news or tech summaries, which increases the chance of accidental or implicit activation of a workflow that performs multiple downstream actions. In this skill, that risk is amplified because the workflow does not stop at summarization; it proceeds toward file writes and publication, so an innocuous request could unintentionally initiate a high-impact content pipeline.

Vague Triggers

Medium

Confidence: 92% confidence
Finding: The usage examples include generic phrases like 'AI日报' and '写一篇技术深度文章' that overlap with normal user intent, making it difficult to distinguish when the full skill should activate. Because this skill chains search, aggregation, and publishing behavior, ambiguous activation creates a real risk of over-executing beyond what the user expected.

Missing User Warnings

High

Confidence: 98% confidence
Finding: The skill includes an external HTTP POST to publish article content to a remote server and describes公众号/飞书 publication without any user-facing warning, consent checkpoint, or disclosure of what data leaves the environment. This is dangerous because collected content, generated text, metadata, or potentially sensitive workspace material could be transmitted or publicly posted unintentionally.

Vague Triggers

Medium

Confidence: 93% confidence
Finding: The trigger list is broad and consists of common technology-related phrases, which increases the chance the workflow is invoked unintentionally during ordinary user conversations. Because this skill performs multi-step external data collection and downstream file/output actions, accidental activation can lead to unnecessary processing, unintended content generation, or follow-on publishing behavior.

Missing User Warnings

Medium

Confidence: 95% confidence
Finding: The workflow explicitly includes saving output to a filesystem path under /root and optionally posting to a public account API or creating an external archive, but there is no manifest-level indication of user confirmation, approval gating, or publication safeguards. In a skill that can be triggered by broad phrases, silent write/publish behavior materially raises the risk of unauthorized local writes, unintended data propagation, or accidental publication of generated content.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.