Back to skill

Security audit

zhongcao-local-business-note-maker

Security checks across malware telemetry and agentic risk

Overview

The skill is mostly aligned with making Xiaohongshu local-business content, but it also grants broad Beatra account authority and silently self-updates local package files by default.

Review this carefully before installing. It appears intended for legitimate Beatra-powered Xiaohongshu content creation, but installation gives a shared Beatra credential broad paid media permissions and the bundled client self-updates by default. Disable auto-updates if you require review before code changes, upload only images you are comfortable sending to Beatra, and confirm Xiaohongshu lookups one at a time because they are paid reads.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Rogue AgentSelf-Modification, Session Persistence
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Findings (17)

Lp3

Medium
Category
MCP Least Privilege
Confidence
94% confidence
Finding
The skill declares no permissions, yet its instructions require shell execution, network access, file reads/writes, environment use, and local client operations. That mismatch hides the real trust boundary from users and reviewers, increasing the chance that sensitive local data, credentials, or files are accessed under the guise of a simple content-generation skill.

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding
The public description frames the skill as a Xiaohongshu note maker, but the body also includes credential handling, generic remote tool invocation, artifact upload, installation registration, uninstall logic, and self-update behavior. This is a substantial behavior mismatch that can mislead users into authorizing infrastructure-management and data-transfer actions they would not expect from a content-planning tool.

Description-Behavior Mismatch

Medium
Confidence
91% confidence
Finding
The skill adds self-updating installation-management behavior unrelated to its stated note-making purpose. Embedding package maintenance into an end-user content skill expands the attack surface and creates an avenue for code changes after initial review, especially if users did not intentionally install a software updater.

Context-Inappropriate Capability

Medium
Confidence
93% confidence
Finding
Automatic package update and file replacement are powerful capabilities not justified by the skill's stated business-content function. Even with integrity-check claims, any mechanism that downloads and replaces local package-owned files materially increases supply-chain and post-installation compromise risk.

Context-Inappropriate Capability

High
Confidence
98% confidence
Finding
The authorization helper requests a very broad OAuth scope set including wallet spending, voice write, music/video generation, and task/artifact permissions that are not necessary for a skill whose stated purpose is generating local-business posts. Over-scoped tokens violate least privilege and materially increase blast radius if the token is misused, stolen, or the skill later invokes capabilities outside user expectations.

Description-Behavior Mismatch

High
Confidence
96% confidence
Finding
The client contains a full self-update subsystem that downloads manifests and archives, mutates the local installation, maintains rollback state, and performs silent auto-update. For a skill whose stated purpose is local-business note creation/research, this is unnecessary privileged behavior and creates a remote code-supply-chain path whereby compromise of the update channel or publisher infrastructure could alter code on the user device.

Description-Behavior Mismatch

Medium
Confidence
94% confidence
Finding
The code records local skill inventory and sends installation registration telemetry, including package slug, version, platform, and external installation reference, despite that behavior not being part of the skill's content-generation purpose. Undisclosed device-local tracking and outbound telemetry increase privacy and monitoring risk, especially when triggered on ordinary operations rather than a dedicated consented registration step.

Context-Inappropriate Capability

Medium
Confidence
90% confidence
Finding
The skill fingerprints its host environment by inspecting process environment variables and persisted host metadata to classify whether it is running under Claude Code, Codex, or another agent platform. That is unrelated to note generation and creates unnecessary environment reconnaissance that can be used for profiling, telemetry enrichment, or platform-specific behavior.

Missing User Warnings

Medium
Confidence
89% confidence
Finding
The skill states that newer releases install automatically without separate confirmation, but this is not prominently disclosed in the top-level description. Silent code changes undermine informed consent and can cause users to run materially different behavior than the version they originally trusted or reviewed.

Missing User Warnings

Medium
Confidence
86% confidence
Finding
The manifest advertises optional reading of platform notes, comments, and account posts but does not clearly disclose privacy boundaries, consent requirements, retention, or how external content will be used. This can mislead users into authorizing collection and analysis of third-party or account data without understanding the data-use implications.

Missing User Warnings

Medium
Confidence
88% confidence
Finding
The English localized description describes reading Xiaohongshu notes, top comments, and recent account posts, but it frames this as a product feature without a clear warning that external content access will occur. In a content-generation skill, this increases the risk of undisclosed collection or analysis of platform and account data during use.

Missing User Warnings

Medium
Confidence
88% confidence
Finding
The Japanese localized description similarly promotes competitor-note and comment analysis but does not explicitly warn about external content retrieval, account-data access, or privacy implications. Users relying on the localized text may not realize they are authorizing collection and analysis of third-party platform data.

Missing User Warnings

Medium
Confidence
89% confidence
Finding
The Chinese localized description states the skill can directly read Xiaohongshu notes, hot comments, and recent account posts for analysis, but it does not provide an explicit warning about privacy, authorization, or downstream use of the data. Given the skill's research and competitor-analysis context, that omission materially increases data-handling risk.

Missing User Warnings

Medium
Confidence
90% confidence
Finding
The document states the client silently checks for updates and automatically installs newer versions by default without separate confirmation. Even with strong integrity checks and rollback protections, automatic file replacement changes local software state without explicit user approval at update time, which creates supply-chain and change-management risk if the update source or signing pipeline is ever compromised.

Missing User Warnings

Medium
Confidence
82% confidence
Finding
The document states that the client automatically performs an installation registration call and writes a local registrations cache, but it does not explicitly warn users that metadata is being recorded and transmitted. Even though the data described is limited and non-secret, silent telemetry and local persistence can create privacy, compliance, and trust issues, especially in enterprise or regulated environments where users expect explicit disclosure and consent.

Missing User Warnings

Medium
Confidence
89% confidence
Finding
The workflow directs users to upload local storefront and brand reference images to a remote service, but it does not include any explicit warning that these files leave the local environment or any guidance about sensitive content in images. In this skill context, users may provide business photos that contain addresses, people, signage, schedules, or other identifying information, so the omission creates a real privacy and data-transfer risk rather than a purely theoretical issue.

Self-Modification

High
Category
Rogue Agent
Content
)
    update = subparsers.add_parser(
        "update",
        help="Check, install, or configure Beatra package self-updates",
    )
    update.add_argument(
        "--check",
Confidence
97% confidence
Finding
The exposed self-update command enables the package to replace its own local code and mutate installation state from remote content. Even with validation checks, self-modifying behavior is high risk for a content-creation skill because it broadens trust from the installed package to ongoing network-controlled release infrastructure and can facilitate persistence or post-install capability changes.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.