Back to skill

Security audit

channels-video-publish

Security checks for vulnerabilities and agentic risk

Overview

This video-publishing skill needs review because it requires disabling execution isolation and can change post descriptions or publish without fully verifying the requested date.

Before unattended use, correct full-date verification and preserve existing descriptions during rescheduling. Retain execution isolation with narrowly scoped browser access and use a dedicated browser profile. The --dry-run option uploads your video and saves a real platform draft; review its content, schedule, and originality declaration.

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
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (12)

Incomplete referenced artifact analysis

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 77)May include surrounding context.

md
- 核查原创:`node scripts/check-original.js`

Incomplete referenced artifact analysis

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 105)May include surrounding context.

md
node scripts/check-original.js [--all]

Incomplete referenced artifact analysis

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 148)May include surrounding context.

md
- `scripts/check-original.js` — 原创声明状态核查

Incomplete referenced artifact analysis

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 93)May include surrounding context.

md
- **对账**:`node scripts/list-all.js` — 一次拉全量列表(标题/时间/原创状态),

Incomplete referenced artifact analysis

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 106)May include surrounding context.

md
node scripts/list-all.js

Incomplete referenced artifact analysis

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 146)May include surrounding context.

md
- `scripts/list-all.js` — 全量列表导出,用于对账「哪些发了哪些没发」

Incomplete referenced artifact analysis

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 168)May include surrounding context.

md
- 新增 `list-all.js`:一次性拉全量列表,用于对账

Skill declares no tool scope ('permissions' or 'allowed-tools') but code capabilities were detected: network.

Medium
Category
MCP Least Privilege
Confidence
70% confidence
Finding

Without declared permissions the skill's intent is opaque and cannot be validated.

Content

No source excerpt is available for this finding.

The prerequisites require disabling the Bash sandbox without warning about the loss of execution isolation.

Medium
Category
Output Handling
Confidence
97% confidence
Finding

The prerequisites explicitly require dangerouslyDisableSandbox: true without explaining the loss of isolation or limiting its use. Running Node scripts and dependencies outside the sandbox can expose host files and network resources if that code is compromised; the authenticated Chrome debugging session also provides a path to account actions. A separate Chrome profile reduces browser-profile mixing but does not restore execution isolation.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · references/troubleshooting.md (reported line 17)May include surrounding context.

Chrome 是按请求的 Host 头生成 webSocketDebuggerUrl 的。 发 Host: 127.0.0.1 会拿到 ws://127.0.0.1/devtools/...(没有端口)→ ws 去连 80 → ECONNREFUSED。 curl 能连上是因为它发的是 Host: 127.0.0.1:9222。

js
http.get({ host: '127.0.0.1', port: 9222, path, headers: { Host: '127.0.0.1:9222' } })

The instructions require disabling the execution sandbox without warning about the loss of security isolation.

Medium
Category
Output Handling
Confidence
98% confidence
Finding

The document explicitly requires running Bash with dangerouslyDisableSandbox: true, removing execution isolation to work around a local networking issue. Publishing scripts would then run with the host user's permissions, increasing the consequences of compromised scripts or dependencies, including access to local files and the authenticated browser session. A dedicated Chrome profile reduces profile mixing but does not replace execution isolation.

Content

No source excerpt is available for this finding.

The documented form-only dry run saves a persistent remote draft.

Medium
Category
Excessive Agency
Confidence
97% confidence
Finding

The --dry-run check occurs only after uploading the video and clicking '保存草稿', so a documented form-filling rehearsal can transfer private content and persist a draft in the authenticated remote account. Although this mode does not click Publish and the destination is the intended publishing platform, those persistent side effects exceed typical dry-run expectations and may expose unreleased content to the platform or other authorized account users.

Content

No source excerpt is available for this finding.