Security audit

douyin-content-intelligence-guaikei

Security checks for vulnerabilities and agentic risk

Overview

The skill is mostly purpose-aligned for Douyin public-data research, but it has broad auto-invocation rules and runtime/token handling that deserve review before installation.

Install only if you want ambiguous short-video/content-planning requests to be eligible for Douyin-specific API calls. Treat generated logs as potentially sensitive research data, clean them up when no longer needed, and be aware that token errors may display provider contact/website text despite the skill claiming it will not do that at runtime.

SkillSpector

Vulnerability Patterns

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

Findings (8)

Vague Triggers

Medium

Confidence: 92% confidence
Finding: The description says the skill should be used even when the user does not explicitly mention Douyin, which broadens invocation beyond clear user intent. That can cause unnecessary access to external data sources and execution of token-authenticated tooling in contexts where the user asked for generic content planning rather than platform-specific collection.

Vague Triggers

Medium

Confidence: 93% confidence
Finding: The 'when to use' section repeats the rule to trigger even if the user never mentions Douyin, reinforcing an overbroad activation policy. In an agent setting, repeated ambiguous triggers increase the chance of unintended tool use, surprising data collection, and disclosure of externally sourced results that the user did not ask for.

Vague Triggers

Low

Confidence: 84% confidence
Finding: The package description uses broad capability language such as public data analysis, competitor analysis, and hotspot monitoring without clear activation boundaries or user-intent constraints. In an agent setting, vague triggering can cause the skill to be invoked for loosely related requests and collect or process platform data more aggressively than the user explicitly intended.

Vague Triggers

Medium

Confidence: 90% confidence
Finding: The natural-language mappings include very broad phrases such as “搜索 / 搜一下 / 找 …” and similar everyday wording that can be triggered inside larger, unrelated requests. In a multi-step agent workflow, this can cause the skill to activate unexpectedly and perform external data collection when the user did not clearly intend to invoke this specific Douyin scraping capability.

Vague Triggers

Medium

Confidence: 95% confidence
Finding: The intent recognition priority rules rely on ambiguous triggers like “热搜 / 热点 / 榜单 / 今天什么火” and “搜索 / 搜一下 / 找 + 关键词,” which overlap with common assistant requests far beyond this skill’s scope. Because the skill metadata explicitly says to use it even when the user does not directly say “抖音,” these rules increase the chance of misrouting generic requests into third-party scraping actions and unnecessary token-backed API use.

Missing User Warnings

Medium

Confidence: 84% confidence
Finding: The CLI persists fetched comment results to a local JSON file automatically, and those results can contain personal data or sensitive user-generated content. Because this happens without explicit user warning, opt-in, or controls over storage location/retention, the tool increases the risk of unintended data retention, leakage via local filesystem access, or accidental inclusion in backups and source repos.

Missing User Warnings

Medium

Confidence: 87% confidence
Finding: The CLI writes the fetched Douyin results to a local JSON file automatically, without any opt-in, warning, or data-minimization step. Because the tool is intended for account scraping, competitor research, and comment/reputation analysis, the output may contain sensitive business intelligence or personal data that persists on disk and can be exposed to other local users, backups, logs, or later compromise.

Missing User Warnings

Medium

Confidence: 92% confidence
Finding: The CLI writes the full search output to a local JSON file, which can persist sensitive or proprietary operational data such as searched keywords, returned content, and execution metadata without explicit user consent or a clear retention policy. In this skill's context, users may be doing competitor research, topic planning, or comment analysis, so silent disk persistence increases the risk of unintended disclosure on shared machines, CI runners, or agent hosts.

Static analysis

No suspicious patterns detected.