Back to skill

Security audit

Google Flow Video Automation

Security checks for vulnerabilities and agentic risk

Overview

The skill is a Google Flow browser automation tool, but it asks users to expose their normal logged-in Chrome profile through Chrome DevTools and automatically confirms credit-spending prompts.

Review before installing. Use a dedicated Chrome automation profile or separate OS account, bind CDP to 127.0.0.1, close the debug browser after use, avoid opening unrelated sensitive sites in that profile, and do not run the scripts unless you are comfortable with them automatically confirming Google Flow credit-spending prompts.

Vulnerability Patterns
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • 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
Findings (2)

T05 · Unauthorized Access and Privilege Escalation

Error
Location
SKILL.md:18
Finding

Primary Authenticated Chrome Profile Exposed Through the DevTools Protocol

Content
View full analysis
Remediation
View remediation

T08 · Insecure Dependencies

Warning
Location
package-lock.json:15
Finding

Dependency Artifact Retrieved from a Third-Party Registry Mirror

Content
View full analysis
Remediation
View remediation
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Memory PoisoningPersistent Context Injection, Context Window Stuffing, Memory Manipulation
  • Tool MisuseTool Parameter Abuse, Chaining Abuse, Unsafe Defaults
  • Rogue AgentSelf-Modification, Session Persistence
Findings (26)

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
85% confidence
Finding

Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous glob patterns).

Content

Scanner excerpt · QUICKSTART.md (reported line 120)May include surrounding context.

解决:

bash
# 删除缓存的 selectors,让脚本重新探索
rm ~/.workbuddy/skills/google-flow-automation/selectors.json

# 重新运行脚本(会重新探索并缓存)
node google-flow-fast.js --prompts prompts.txt

Tp4

High
Category
MCP Tool Poisoning
Confidence
94% confidence
Finding

The description presents an automation skill whose main function is creating AI videos in Google Flow through Chrome CDP with specific output settings and download behavior. This code chunk instead serves as a preparatory connectivity check for a local Chrome debugging endpoint and provides user instructions for launching Chrome. While this may be a supporting utility for a larger automation workflow, the actual behavior of the supplied code does not match the declared primary purpose or listed capabilities. Additionally, the comments mention checking whether the user is logged into Google Flow, but the implementation only queries /json/version on localhost and never verifies Google Flow session state.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
96% confidence
Finding

The code is a first-run exploratory/setup script, not a video generation script. It connects to an existing Chrome instance, prompts the user to manually log in, navigates the Google Flow UI, opens settings, enumerates visible options, saves screenshots, and writes an available-options.json file. There is no prompt submission for generation, no monitoring for completed renders, and no file download logic for MP4 output. Therefore the declared description materially overstates and misrepresents the actual behavior.

Content

No source excerpt is available for this finding.

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 75)May include surrounding context.

md
**脚本会自动处理:** 如果提示词不是以视频关键词开头,`generate-one.js` 会自动添加 `"Video of "` 前缀。

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 177)May include surrounding context.

md
**脚本会自动处理:** 如果提示词不是以视频关键词开头,`generate-one.js` 会自动添加 `"Video of "` 前缀。

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 182)May include surrounding context.

md
**脚本会自动处理:** 如果提示词不是以视频关键词开头,`generate-one.js` 会自动添加 `"Video of "` 前缀。

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 182)May include surrounding context.

md
- ✅ 新增 `explore-options.js` 和 `generate-one.js`

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The quickstart instructs users to force-kill all Chrome processes with pkill -9 but does not clearly warn about loss of unsaved browser work or interruption of active sessions. This is a destructive operational step that can cause immediate user harm even if it is not a traditional exploit.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The guide tells users to launch Chrome with remote debugging against their real profile directory without warning that any local process able to reach the CDP port can control the browser and access authenticated session data. Reusing the default profile magnifies risk because cookies, account state, and browsing data are all exposed to automation.

Content

No source excerpt is available for this finding.

Session Persistence

Medium
Category
Rogue Agent
Confidence
90% confidence
Finding

This duplicate finding identifies the same risky behavior: preserving cookies and keeping an authenticated browser open for reuse. In a CDP-driven automation skill, that practice materially increases exposure of the user's account and session tokens.

Content

Scanner excerpt · QUICKSTART.md (reported line 23)May include surrounding context.

md
### 步骤2:手动登录(一次性)

在打开的 Chrome 窗口中:
1. 点击 **"Create with Google Flow"**
2. 登录你的 Google 账号(只需首次登录,之后 Cookie 保持)
3. 完成登录后,**保持 Chrome 开着**

Session Persistence

Medium
Category
Rogue Agent
Confidence
90% confidence
Finding

This duplicate finding identifies the same risky behavior: preserving cookies and keeping an authenticated browser open for reuse. In a CDP-driven automation skill, that practice materially increases exposure of the user's account and session tokens.

Content

Scanner excerpt · QUICKSTART.md (reported line 23)May include surrounding context.

md
### 步骤2:手动登录(一次性)

在打开的 Chrome 窗口中:
1. 点击 **"Create with Google Flow"**
2. 登录你的 Google 账号(只需首次登录,之后 Cookie 保持)
3. 完成登录后,**保持 Chrome 开着**

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The documentation explicitly frames the approach as a way to bypass Google login/bot-detection by driving a real Chrome profile with persistent authenticated state. That expands the skill from ordinary UI automation into account/session abuse territory and normalizes techniques that can evade platform defenses.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

Planned support for rotating multiple Google accounts is not necessary for simple video generation and is commonly associated with quota evasion, account farming, or abuse of service restrictions. In this context it signals future expansion toward scaling actions across identities rather than a single user's normal workflow.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
87% confidence
Finding

The skill instructs users to launch Chrome with a remote debugging port exposed and relies on network-capable automation, but the manifest does not declare any explicit tool scope or permissions. Undeclared network/browser-control capability reduces review transparency and can hide risky behavior such as browser session access, especially because CDP can expose authenticated browser state.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The main instructional content is written entirely in Chinese, while the metadata description is in English, creating a language constraint for users who need to follow the skill instructions. Under the policy, forcing a specific language without user opt-in or a documented regional justification is a natural-language policy violation.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

This JavaScript file contains user-facing comments and console messages entirely in Chinese, including the operational guidance shown to the user. The policy requires avoiding forced language/locale constraints unless the skill offers opt-in choice or clearly documents a justified regional limitation, which is not present here.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

Line L06 states that using English prompts works better, and the file consistently instructs prompt construction around that expectation without offering any user opt-in or alternative language handling. This creates a natural-language locale policy issue because it steers users toward a specific language by default rather than presenting language choice or a documented, justified regional constraint.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The file’s natural-language interface, usage guidance, and runtime prompts are entirely in Chinese, including the onboarding text and user instructions. This creates a language/locale policy issue because the skill does not offer user opt-in or an alternative language, and the file does not document that it is intentionally region-specific.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The script automatically detects credit/cost confirmation prompts and clicks "Yes" without any explicit user approval at runtime. In this skill's context, that behavior can trigger billable video-generation actions on the user's logged-in Google Flow account, converting a browser automation task into an unauthorized spending risk.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

This JavaScript file includes its primary description, usage guidance, and many operational comments in Chinese, which imposes a specific language on users without any opt-in or alternative locale. The stated policy requires flagging language or locale constraints unless the skill explicitly offers a language choice or clearly documents a justified regional scope.

Content

No source excerpt is available for this finding.

Context Window Stuffing

Medium
Category
Memory Poisoning
Confidence
85% confidence
Finding

Skill attempts to fill the context window with filler content, displacing legitimate instructions and safety constraints. This can degrade agent performance or bypass safety boundaries.

Content

Scanner excerpt · google-flow-automate.js (reported line 135)May include surrounding context.

js
}, promptText);
  
  if (!result.success) {
    throw new Error('Failed to fill prompt: ' + result.error);
  }
  
  console.log(`✅ Prompt filled using: ${result.selector}`);

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

This code fetches video content from a remote URL and saves it to a local output path, which are safety-relevant network and file-write operations. Although it logs that a download is happening, the file does not provide a meaningful user-facing warning about remote data transfer or local persistence of generated files beyond routine status output.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
78% confidence
Finding

The phrase 提示词建议(英文效果更佳) communicates a preferred language for use of the skill. Although mild, it can be read as a language-policy bias without explicitly making clear that other languages remain supported and that English is only an optional optimization tip.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
88% confidence
Finding

The file's header comments, usage text, and most user-facing console output are in Chinese, with no indication that other languages are supported. This creates a language/locale policy issue because the skill imposes a specific language experience without offering the user a choice.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
97% confidence
Finding

The human-readable documentation and comments are written entirely in Chinese, including the usage guidance, with no indication of language choice or opt-in. This can violate language/locale policy when a skill imposes a specific language on users without providing an option.

Content

No source excerpt is available for this finding.