Back to skill

Security audit

小红书数据助手

Security checks for vulnerabilities and agentic risk

Overview

ClawScan rates this skill suspicious because scanners report sensitive information being sent to a third party and compliance records that claim confirmation or review without verifying it.

Consider the reported credential and token disclosures, unpinned installer, and unsupported compliance records before installing. The scan is partial: although all listed components were inspected, some references could not be resolved, so the evidence is incomplete. The verdict is suspicious, not a claim that the author acted maliciously.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Vulnerability Patterns
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Findings (15)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

Local bookkeeping can be a supporting implementation detail, but this code does more than log an operation: it records confirmation as an accomplished fact without any consent check in the supplied chunk. That is materially different from the declared analysis functionality and confirmation safeguards. The mismatch is the unsupported creation of confirmation records, not merely the absence of API calls in this excerpt.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
93% confidence
Finding

Ordinary operational logging could be a supporting implementation detail, but this code specifically records a completed legal review without performing or validating one. That compliance-status assertion is not disclosed by the declared data-analysis purpose. The mismatch is the unsupported review-completion record, not merely the existence of a local log.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
91% confidence
Finding

Local compliance storage and audit logging could be supporting implementation details. However, this code does more than log activity: it unconditionally marks legal confirmation as completed. The supplied chunk contains no prompt, consent check, or conditional gate establishing that confirmation occurred. That consent-state mutation is not disclosed by the analytics description and is inconsistent with its emphasis on explicit confirmation. The chunk does not itself perform PGY queries or establish that the stored legal confirmation would authorize them.

Content

No source excerpt is available for this finding.

Pagination instructions recommend forwarding a sensitive login credential to a third-party API without warning the user about credential disclosure.

High
Category
Output Handling
Confidence
94% confidence
Finding

The pagination instructions recommend forwarding session_id while explicitly describing it as a sensitive login credential, yet classify the endpoint as low risk and provide no local credential-disclosure warning. If it is an authentication-bearing session rather than an opaque pagination identifier, forwarding it to the third-party service could enable account access, and GET query parameters can additionally expose it through URL logging.

Content

No source excerpt is available for this finding.

The consent gate requires Chinese confirmation wording without offering a language choice.

Medium
Category
Output Handling
Confidence
91% confidence
Finding

The accepted confirmation phrases are exclusively Chinese, while English responses such as 'yes' are explicitly rejected. This imposes a language restriction without user opt-in or an explanation of why consent must be expressed in Chinese; requiring informed consent does not itself require a particular language.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'npx skillhub'.

Medium
Category
MCP Rug Pull
Confidence
89% confidence
Finding

The installation command can download and execute an unpinned npm CLI, exposing the user's environment to a compromised or unexpectedly changed package release. This is a conditional supply-chain vulnerability, although the finding incorrectly describes the CLI as an MCP server.

Content

No source excerpt is available for this finding.

The consent gate requires Chinese-language confirmation without offering a language choice.

Medium
Category
Output Handling
Confidence
96% confidence
Finding

The accepted confirmations are exclusively Chinese, while English "yes" is explicitly rejected and users are instructed to use the prescribed wording before proceeding. This imposes a language requirement without user opt-in or a documented justification, even though the manifest lists English as a supported locale.

Content

No source excerpt is available for this finding.

The lookup workflows send user-provided share links and security tokens to a third-party API without a privacy warning.

Medium
Category
Output Handling
Confidence
93% confidence
Finding

L031–L032 prescribe passing share_text and xsec_token directly into lookup requests, whose destination is https://www.aconfig.cn (L003), rather than Xiaohongshu itself. Although the endpoint and parameters are documented, the file does not warn users that their supplied links and associated security tokens will be disclosed to this intermediary.

Content

No source excerpt is available for this finding.

The generic trigger “打开链接” (“open link”) can activate the note-detail recipe for unrelated links.

Medium
Category
Trigger Abuse
Confidence
94% confidence
Finding

“打开链接” is a common request that does not identify a note or supported platform, yet it routes to share_link_to_detail. The file provides no URL constraints or exclusion conditions for this trigger; longest-match selection and tie clarification do not prevent unrelated link-opening requests from matching it.

Content

No source excerpt is available for this finding.

The standalone trigger “推荐” (“recommend”) can activate the home-feed workflow for unrelated requests.

Medium
Category
Trigger Abuse
Confidence
99% confidence
Finding

“推荐” is common in requests for products, books, restaurants, and other suggestions. This trigger does not require home-feed or note-browsing intent, yet the workflow fetches the home feed and note details.

Content

No source excerpt is available for this finding.

The confirmation policy requires Chinese wording without offering a language choice.

Medium
Category
Output Handling
Confidence
97% confidence
Finding

All accepted confirmation phrases are Chinese, while English responses “yes” and “ok” are explicitly rejected. This makes continuation dependent on a particular language without user opt-in or a documented justification for that language restriction.

Content

No source excerpt is available for this finding.

The whitelist permits querying the MaxHub account balance, outside the stated Xiaohongshu analytics scope.

Low
Category
Excessive Agency
Confidence
90% confidence
Finding

L365–L371 authorize GET /api/maxhub/balance, which accesses MaxHub account information rather than Xiaohongshu content or creator analytics. Account-balance inspection is neither declared in the manifest nor a direct requirement of the described analytics workflows.

Content

No source excerpt is available for this finding.

The standalone trigger “作者” (“author”) is too broad for a note-author profile workflow.

Low
Category
Trigger Abuse
Confidence
96% confidence
Finding

“作者” can occur in ordinary questions about books, articles, or authorship without requesting a note author's profile. The activation condition does not require an explicit reference to a selected note or exclude unrelated author queries.

Content

No source excerpt is available for this finding.

The trigger “笔记详情” activates two workflows without a routing condition.

Low
Category
Trigger Abuse
Confidence
98% confidence
Finding

The same trigger is listed at L003 for keyword search followed by note details, while this recipe first retrieves a user's posted notes using user_id. No activation rule distinguishes these paths, leaving the intended workflow ambiguous.

Content

No source excerpt is available for this finding.

Risk-reminder triggers include ordinary words without requiring the corresponding risky action.

Low
Category
Trigger Abuse
Confidence
90% confidence
Finding

The batch-operation keywords include “全部” and “所有” (“all”), while personal-data keywords include “笔记” (“notes/posts”) and “种草” (“product recommendations”). Even within the Xiaohongshu context, benign requests can contain these words without requesting bulk collection or personal information, and the configuration specifies no exclusions or required action-keyword combinations.

Content

No source excerpt is available for this finding.