guaikei-douyin-audience-voice-reader

• Outcome: Review
• Latest audit: Aug 24, 2026 · 2:49 AM UTC
• Version: 1.0.0

Overview

The skill mostly matches its public Douyin research purpose, but it has Review-worthy privacy and scope issues around broad activation, automatic local logging, and returned media download URLs.

Install only if you are comfortable sending Douyin research keywords, target URLs, and a GUAIKEI_API_TOKEN to guaikei.com, and with results being saved automatically under the skill's logs directory. Avoid using it for sensitive competitor research unless you manage or delete those logs, and do not treat returned media URLs as authorization to download or redistribute videos.

SkillSpector

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• MCP Tool Poisoning: Hidden Instructions, Unicode Deception, Parameter Description Injection
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access

Findings (8)

Description-Behavior Mismatch

Medium

• Confidence: 89% confidence
• Finding: The schema explicitly documents `video_url` as including both playback and download addresses, which conflicts with the skill metadata stating the skill is not for video downloading. That mismatch can enable downstream agents or integrators to expose or rely on download functionality that the declared scope forbids, increasing the chance of policy bypass and unauthorized content retrieval.

Description-Behavior Mismatch

Medium

• Confidence: 95% confidence
• Finding: The schema explicitly documents `play_addr` as both a playback URL and a video download URL, which conflicts with the skill description stating that downloading is out of scope. This mismatch can enable downstream agents or integrators to use the field for prohibited content downloading, expanding the skill's effective capability beyond its declared boundaries and increasing compliance and misuse risk.

Description-Behavior Mismatch

Medium

• Confidence: 92% confidence
• Finding: The CLI writes fetched comment data to a local JSON file after completing a read operation, which exceeds a purely read-only audience inspection behavior and creates an undisclosed persistence channel. Even though the data appears to be public comments, local storage increases privacy, retention, and multi-user exposure risk on shared systems and broadens the skill's effective capabilities beyond what the manifest describes.

Vague Triggers

Medium

• Confidence: 86% confidence
• Finding: The trigger guidance says the skill should activate even when the user did not mention Douyin, including generic 'short-video research' or multi-step requests. That can cause the agent to send user queries to an external data-collection workflow without clear platform-specific user intent, increasing the risk of unintended data disclosure and over-collection.

Missing User Warnings

Medium

• Confidence: 92% confidence
• Finding: The skill states that output files are automatically saved to `logs/` with timestamps and query parameters, but the user-facing description does not clearly warn that their search terms and requested URLs may be persisted. This creates a privacy and retention risk because sensitive research topics, competitor targets, or user-provided links could be stored on disk without explicit notice or consent.

Missing User Warnings

Medium

• Confidence: 90% confidence
• Finding: The tool silently saves retrieved comments to local disk without explicit user notice in this file. Silent persistence can surprise users, leak collected data to other local users or processes, and leave residual files that outlive the intended session.

Missing User Warnings

Medium

• Confidence: 82% confidence
• Finding: The CLI sends the supplied URL and request parameters to an external API to fetch comments, but this network transmission is not clearly disclosed in the analyzed file. Because user-provided inputs may contain sensitive targets, internal links, or monitoring-sensitive queries, undisclosed outbound transmission expands data exposure and trust boundaries.

Missing User Warnings

Medium

• Confidence: 93% confidence
• Finding: The CLI persists the full search output to a local file named with the user-provided keyword, without any explicit consent, warning, or opt-out. In this skill context, search queries and returned public-platform content can still be sensitive from the operator’s perspective, and writing them to disk increases exposure through local file access, backups, shared workstations, or later collection by other tooling.

VirusTotal

VirusTotal findings are pending for this skill version.

Static analysis

No suspicious patterns detected.