Security audit

X(Twitter)作品搜索

Security checks across malware telemetry and agentic risk

Overview

This skill is a disclosed X/Twitter search helper that uses a third-party API and API key without hidden persistence or destructive behavior.

Before installing, understand that your X/Twitter search terms and pagination data will be sent to redfox.hk using your Redfox API key, and results may include public author profile details, locations, metrics, and media URLs. Avoid sensitive searches if third-party processing is not acceptable for your use case.

SkillSpector

By NVIDIA

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access

Findings (7)

Context-Inappropriate Capability

Medium

Confidence: 91% confidence
Finding: The workflow instructs the agent to discover credentials from both an environment variable and a local user-home config file. Even if intended for normal authentication, this expands secret-access scope beyond what is justified in the visible skill context and conditions the agent to inspect local secret locations, which can normalize unnecessary sensitive-file access and increase the chance of accidental disclosure or misuse.

Vague Triggers

Medium

Confidence: 87% confidence
Finding: The skill advertises activation through broad natural-language requests such as 'Just describe what you need in natural language,' which can cause the skill to trigger on loosely related user utterances rather than explicit invocation. In an agent environment with multiple tools, this raises the risk of unintended tool execution, unexpected external queries, and user confusion about what action is being taken.

Vague Triggers

Medium

Confidence: 93% confidence
Finding: The pagination triggers include very generic phrases like 'Next page,' 'Continue,' and 'More,' which are common conversational utterances that may appear in unrelated contexts. This makes accidental re-invocation likely, potentially causing repeated external requests, unintended data retrieval, and loss of user control over when the skill continues operating.

Vague Triggers

Medium

Confidence: 88% confidence
Finding: The README defines activation through very broad natural-language phrases like '搜一下', '看看', '继续', and '还有吗' without clear boundaries or disambiguation rules. In an agent environment, this can cause accidental invocation or unintended continuation of prior searches, leading to unrequested retrieval of external content and potential over-collection of data.

Missing User Warnings

Medium

Confidence: 84% confidence
Finding: The description advertises extraction of tweet text, engagement metrics, media attachments, and author information, but does not clearly warn users that profile data and media from X will be collected and returned. This weakens informed consent and increases privacy/compliance risk, especially when users may not expect author metadata, locations, or media URLs to be surfaced in bulk.

Vague Triggers

Medium

Confidence: 92% confidence
Finding: The skill’s trigger phrases are broad and overlap with common user intents such as general 'X/Twitter search' requests, which can cause the skill to activate when the user did not specifically intend to invoke this tool. Unintended activation can route user queries to an external API and expose search terms or conversation context unnecessarily, especially because the skill is designed to fetch third-party platform data automatically.

Missing User Warnings

Medium

Confidence: 89% confidence
Finding: The workflow sends user-provided search terms, pagination state, and an identifying source tag to an external third-party API, but the skill content does not provide an up-front privacy/transmission warning or obtain clear user acknowledgment. Search queries can contain sensitive interests, names, or other personal data, so silent transmission to an external service creates a privacy risk and may violate user expectations or organizational data-handling rules.

VirusTotal

63/63 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.