Back to skill

Security audit

小红书自动发布

Security checks for vulnerabilities and agentic risk

Overview

The skill matches its stated goal of auto-posting to Xiaohongshu, but it can publish from a logged-in account without a final confirmation and exposes the user's main Chrome session through a fixed debugging port.

Install only if you are comfortable with the skill using your logged-in Chrome session to publish real Xiaohongshu posts automatically. Before use, narrow the trigger, add a final confirmation or draft mode, use an isolated browser profile instead of the primary Chrome debugging port, verify the exact media/text files, and remove or protect diagnostic screenshots.

Vulnerability Patterns
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • 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
Findings (2)

T09 · Insecure Skill Coding Practices

Warning
Location
skill.js:207
Finding

Predictable screenshot files in a shared temporary directory

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Error
Location
skill.js:9
Finding

Authenticated primary browser exposed through a fixed Chrome DevTools Protocol port

Content
View full analysis
{ if (code === 0) { console.log('✅ Chrome 调试模式已在运行'); resolve(); } else { // 启动 Chrome console.log('🔄 启动 Chrome...'); const chromePath = '/Applications/Google Chrome.app/Contents/MacOS/Google Chrome'; const userDataDir = process.env.HOME + '/Library/Application Support/Google/Chrome'; const chrome = spawn('open', ['-a', 'Google Chrome', '--args', '--remote-debugging-port=9222', '--no-first-run', '--no-default-browser-check' ], { ``` ```js const browser = await puppeteer.connect({ browserURL: 'http://localhost:9222' }); ``` ### Technical Analysis The skill starts or reuses Chrome with the Chrome DevTools Protocol exposed on the fixed port `9222`. It then connects Puppeteer to the entire browser instance. The code does not launch a dedicated temporary automation profile, use an unpredictable endpoint, authenticate the debugging connection, or terminate an isolated browser after publication. Because the skill invokes the normal installed Chrome application, it may expose a browser containing the user's existing authenticated sessions. A process that can connect to the debugging endpoint can potentially enumerate pages, read page content, execute JavaScript in browser contexts, navigate tabs, and perform actions using active authenticated sessions. The endpoint check only verifies that a service responds successfully through `curl`; it does not validate that the endpoint belongs to the Chrome instance expected by the skill. Consequently, an existing local service on port 9222 could also be trusted as the Puppeteer target. Although loopback access normally limits exploitation to local process ...[truncated 1874 chars]
Remediation
View remediation
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (8)

Missing User Warnings

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

The script automatically triggers the final '发布' action on a live social media platform using an already-authenticated Chrome session, with no confirmation prompt, dry-run mode, or human approval gate. Because publishing is irreversible or difficult to fully undo once content is distributed, accidental execution, misuse by another local process, or unexpected page matching could cause unintended public posts from the user's account.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill explicitly automates posting content to a live external social platform, but the description does not warn the user that invoking the trigger will cause a real publication action. This creates a meaningful risk of unintended posting, reputational harm, and accidental disclosure of content if the user does not understand that the action is not a dry run or draft-only workflow.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The script forces a Chinese title string and later depends on Chinese UI text such as '标题' and '发布', indicating a fixed locale assumption. There is no opt-in or documented justification for restricting operation to a Chinese-language environment.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The script reads local files from a fixed directory and uploads the video while also sending text content from a local file to the remote platform. Although this appears to be the skill's intended function, it still creates a data exfiltration risk because local content is transmitted off-device without an explicit disclosure, consent checkpoint, or validation that the selected files are the intended ones.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The manifest explicitly advertises automatic publishing to a social-media platform but provides no warning that the skill can perform irreversible, account-impacting actions. In this context, lack of disclosure and confirmation materially increases the risk of accidental posting, reputational harm, and misuse of an authenticated browser session.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The trigger phrase "发布笔记" is generic enough to match ordinary user requests that may not specifically intend Xiaohongshu publishing. In a skill that performs account-affecting actions, broad activation increases the chance of accidental invocation and unintended posting to a live social-media account.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The requirements mention browser cookies for YouTube downloads without any warning about handling sensitive authenticated session material. Cookie use can expose account tokens or extend trust from an already logged-in browser, making compromise or misuse more dangerous if the skill accesses, stores, or forwards those credentials insecurely.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
84% confidence
Finding

All user-facing instructions, trigger phrases, and naming are presented only in Chinese, and the usage example requires the Chinese phrase “发小红书” to activate the skill. Under the policy, forcing a specific language without opt-in can be a natural-language policy violation unless the locale constraint is explicitly justified.

Content

No source excerpt is available for this finding.