Security audit

guaikei-douyin-omni-public-data

Security checks across malware telemetry and agentic risk

Overview

This is a coherent Douyin public-data collection skill, but its broad auto-activation rules and automatic local result logging make it something users should review before installing.

Install only if you are comfortable with guaikei.com receiving your Douyin research queries, target URLs, and API token, and with fetched public data being saved locally under logs by default. Use it only when you explicitly want Douyin public-data research, and delete or protect generated logs if the research targets, comments, or account data are sensitive in your context.

SkillSpector

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
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
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access

Findings (8)

Description-Behavior Mismatch

Medium

Confidence: 95% confidence
Finding: The schema explicitly describes `play_addr` as both a playback address and a video download address, which contradicts the skill metadata stating it does not handle video publishing or downloading. In a data-collection skill for Douyin analysis, exposing or normalizing download URLs can enable unauthorized content acquisition, policy bypass, or downstream misuse by agents that trust the schema more than the manifest.

Vague Triggers

High

Confidence: 94% confidence
Finding: The skill description says it should be used even when the user does not mention Douyin, for broad requests like competitor analysis or 'what is hot in the industry.' This can cause overbroad activation in unrelated research workflows, leading the agent to call an external data-collection skill and process platform-specific data without clear user intent or platform scoping.

Vague Triggers

High

Confidence: 95% confidence
Finding: The generalized activation rule explicitly allows triggering for broad multi-step research tasks where Douyin is not named. In an agent setting, this increases the risk of unintended tool use, data overcollection, and confusing cross-platform substitution, especially because the same section later says other platforms should not trigger this skill.

Missing User Warnings

Medium

Confidence: 80% confidence
Finding: The README states that data will be automatically exported as JSON logs, but it does not prominently warn that collected public-platform data and analysis inputs may persist to local disk. This can create unintended data retention and secondary exposure risk, especially if comments, account data, or research targets are sensitive in the user's context or if logs are later shared or synced.

Missing User Warnings

Low

Confidence: 69% confidence
Finding: The skill instructs users to set GUAIKEI_API_TOKEN in an environment variable but does not clearly explain that the runtime reads credentials from the environment or warn against exposing them through shell history, shared terminals, screenshots, or inherited process environments. This is a mild credential-handling weakness in documentation rather than direct secret exfiltration.

Vague Triggers

Medium

Confidence: 86% confidence
Finding: The natural-language invocation mapping includes very broad phrases such as '搜索', '找', and similar everyday wording that could cause the agent to invoke this data-collection skill unexpectedly. In an agent environment, overbroad triggers can lead to unintended external queries, data collection, token use, and local logging without sufficiently explicit user intent.

Vague Triggers

Medium

Confidence: 90% confidence
Finding: The intent-recognition rules prioritize broad phrases like '热点', '搜索', and '找 + 关键词' without strong activation boundaries, which increases the chance of accidental tool invocation from ordinary conversation. Because this skill performs external data access and writes logs, ambiguous routing is more dangerous here than in a purely local or read-only helper skill.

Missing User Warnings

Medium

Confidence: 87% confidence
Finding: The CLI writes the full fetched results to a local JSON file automatically after successful execution, without any explicit user confirmation, opt-in flag, or warning. Because the results may contain scraped account and content data, this creates an unnecessary local data-retention risk: sensitive or regulated data can persist on disk, be read by other local users/processes, or be collected unintentionally from shared environments and CI runners.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.