Security audit

guaikei-xhs-kol-check

Security checks across malware telemetry and agentic risk

Overview

This skill coherently fetches public Xiaohongshu data through a disclosed third-party API, with local result logging that users should understand before use.

Install only if you are comfortable sending Xiaohongshu keywords, note/profile URLs, and API token-authenticated requests to Guaikei, and with fetched results being saved locally in logs. Avoid submitting private, non-public, sensitive business targets unless authorized, and delete logs when no longer needed.

SkillSpector

By NVIDIA

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 (7)

Description-Behavior Mismatch

Medium

Confidence: 88% confidence
Finding: The README presents the skill as a limited content-topic/reference tool, but the actual description expands into competitive monitoring, KOL screening, and trend prediction. This mismatch can cause users or downstream policy systems to under-assess the scope of data collection and surveillance-style use cases, leading to inappropriate deployment or trust decisions.

Description-Behavior Mismatch

Medium

Confidence: 91% confidence
Finding: The CLI writes the full fetched result set to a local JSON log file, including the requested profile URL and returned content data, without any clear retention policy, consent flow, or minimization. For a skill presented as a data-reference tool, silent persistence increases privacy and data-handling risk because scraped or user-requested content may accumulate on disk and be accessible to other local users, backups, or later processes.

Description-Behavior Mismatch

Medium

Confidence: 95% confidence
Finding: The CLI persists the full search output to a local JSON file, including the user's query and all returned results, without any user-facing disclosure or opt-in. In a skill positioned as providing reference data for content ideation, this creates unnecessary data retention and increases the risk of exposing potentially sensitive research topics or scraped platform content through local files.

Missing User Warnings

Medium

Confidence: 89% confidence
Finding: The skill routes user-supplied Xiaohongshu links, keywords, and potentially returned content through a third-party API, but the privacy warning is buried near the end under compliance notes rather than presented clearly before use. Users may provide personal, business-sensitive, or account-linked URLs without realizing that this data is being transmitted externally, creating consent, confidentiality, and data-governance risk.

Missing User Warnings

Medium

Confidence: 90% confidence
Finding: Automatically saving all task results to the logs directory without clearly warning that fetched content, profile links, note URLs, comments, or analysis targets may be written to disk creates a data persistence risk. Local logs can expose scraped public data, operational targets, or sensitive business research to other users, backups, or later compromise.

Missing User Warnings

Medium

Confidence: 83% confidence
Finding: The command transmits the supplied Xiaohongshu profile URL and request parameters to a remote API and later writes results locally, but the user-facing flow does not provide an explicit warning or consent step about outbound transmission and local storage. This is dangerous because users may assume the tool only performs local reference lookups, while in reality it discloses request data to external services and retains output on disk.

Missing User Warnings

Medium

Confidence: 96% confidence
Finding: Writing search results to a local file without warning or consent is a privacy and transparency issue. The output contains the user's keyword and returned data, so local persistence can unintentionally create an audit trail of user interests or collected content that other users or processes on the system may access.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

Detected: suspicious.exposed_secret_literal

File appears to expose a hardcoded API secret or token.

Critical

Code: suspicious.exposed_secret_literal
Location: SKILL.md:16