小红书视频发布

• Outcome: Review
• Latest audit: Sep 11, 2026 · 3:49 AM UTC
• Version: 1.0.0

Overview

This skill automates publishing to a live Xiaohongshu account, but it forces a specific location and does not require explicit final approval before posting.

Install only if you are comfortable with an agent using your logged-in Xiaohongshu creator session. Before use, remove or override the fixed '苏州中心' location unless it is accurate, and require a final confirmation of the title, body, video, cover, location, and account before publishing.

A.I.G

Vulnerability Patterns

• Skill Instruction Hijacking: Alters the agent's session goals or safety constraints when the skill loads
• Agent Memory Poisoning: Writes attacker-controlled rules into memory that affect later sessions
• Remote Payload Retrieval and Execution: Fetches external code whose behavior can change after review
• Embedded Malicious Code: Ships malicious scripts inside the skill and executes them locally
• Unauthorized Access and Privilege Escalation: Obtains permissions beyond the task's legitimate needs

Findings (1)

other

Warning

• Location: SKILL.md:37

Finding

Hardcoded Location Metadata Added Without User Confirmation

Content

View full analysis

Vulnerability Details

File Location: SKILL.md, lines 37–40
Vulnerability Type: Unauthorized content metadata manipulation
Risk Level: Medium

Relevant excerpt translated into English:

markdown

### 6. Add a location
- Click "Add location"
- Search for "Suzhou Center" and select it
- This may increase regional exposure

Technical Analysis

The skill unconditionally instructs the agent to associate every published video with the fixed location “Suzhou Center.” The location is not derived from user input, video metadata, or another validated source, and the workflow does not require the user to confirm it.

Adding this metadata is unnecessary for the skill's primary purpose of publishing a video. Because the workflow later instructs the agent to publish the post, the hardcoded location can become publicly visible or otherwise associated with the user's content and account.

This issue does not provide operating-system privileges, account takeover, or arbitrary code execution. Its scope is limited to unauthorized modification of post metadata within the already authenticated Xiaohongshu publishing session.

Attack Path

1. A user invokes the skill to publish a video on Xiaohongshu.
2. The agent accesses the creator portal using the user's authenticated browser session.
3. The agent uploads the video and fills in the post details.
4. The hardcoded instruction causes the agent to search for and select “Suzhou Center.”
5. The agent publishes the post without separately asking whether the location is accurate or authorized.
6. The resulting post contains potentially false or unwanted location metadata.

Impact Assessment

The issue may:

• Misrepresent the location of the user, video, business, or depicted event.
• Create an unintended public association between the user's account and a specific place.
• Affect regional targeting, discoverability, or platform recommendations.

...[truncated 400 chars]

Remediation

View remediation

Remediation Suggestions

1. Remove the fixed instruction to select “Suzhou Center.”
2. Leave location metadata unset by default because it is not required to publish a video.
3. Add a location only when the user explicitly provides one or confirms a proposed location.
4. Before publishing, show the selected title, description, tags, location, and other externally visible metadata for user approval.
5. Validate that any selected location matches the user's instruction rather than inferring it solely for engagement optimization.
6. If no location was requested, explicitly verify that the location field remains empty during the final pre-publication check.
7. Preserve the existing duplicate-publication safeguards while adding a confirmation boundary immediately before the irreversible publish action.

SkillSpector

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access
• Supply Chain: Unpinned Dependencies, External Script Fetching, Obfuscated Code
• Excessive Agency: Unrestricted Tool Access, Autonomous Decision Making, Scope Creep

Findings (2)

Natural-Language Policy Violations

Medium

• Category: Not specified by scanner
• Confidence: 93% confidence

Finding

The skill hard-codes selection of the location "苏州中心" without asking the user or explaining why that location should be used. This can misrepresent the post's origin, violate user intent, and leak or fabricate location metadata on a public platform.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium

• Category: Not specified by scanner
• Confidence: 95% confidence

Finding

The skill instructs the agent to click a live platform's "发布" button and verify the post afterward, but it does not require an explicit confirmation immediately before the irreversible external action. In an automation context, this creates a real risk of accidental publication of unintended, incorrect, or sensitive content to a public account.

Content

No source excerpt is available for this finding.