Back to skill

Security audit

抖音作品查询

Security checks for vulnerabilities and agentic risk

Overview

The search functionality is coherent, but it can automatically install payment software and initiate a recharge without a separate approval step.

Before installing, require separate approval for payment-software installation and each recharge, including the merchant and total amount. Use a reviewed, pinned payment-tool version and validate payment destinations. Search keywords go to RedFox and consume credits; enable daily searches only if you want recurring usage. The artifacts support review concerns about automatic setup and payment initiation, rather than credential theft or proven unauthorized withdrawals.

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
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Taint TrackingDirect Taint Flow, Variable-Mediated Taint Flow, Credential Exfiltration Chain
Findings (11)

Tainted flow: 'req' from os.environ.get (line 102, credential/environment) → urllib.request.urlopen (network output)

Critical
Category
Data Flow
Confidence
90% confidence
Finding

Credentials or environment variables flow to a network sink. This is a high-confidence indicator of credential exfiltration.

Content

Scanner excerpt · scripts/search_douyin.py (reported line 114)May include surrounding context.

python
)

    try:
        with urllib.request.urlopen(req, timeout=15) as resp:
            result = json.loads(resp.read().decode("utf-8"))
    except urllib.error.HTTPError as e:
        body = e.read().decode("utf-8", errors="replace")

The search workflow escalates an insufficient-credit response into an automatic paid recharge without requiring explicit purchase authorization in the visible instructions.

High
Category
Prompt Injection
Confidence
93% confidence
Finding

The visible workflow converts an insufficient-credit response into payment submission and automatic search resumption without an explicit purchase-authorization gate. Authorizing a search does not authorize spending money, and allowing an API response containing payment details to initiate this escalation creates a financial trust-boundary risk. The referenced payment instructions are unavailable, so downstream safeguards cannot be verified.

Content

No source excerpt is available for this finding.

Search-error output instructs the agent to initiate a payment without an explicit user-approval step.

High
Category
Prompt Injection
Confidence
97% confidence
Finding

An insufficient-credit response causes the script to print instructions directing the agent to initiate an Alipay workflow using API-supplied payment links, amounts, and merchant information, without requiring explicit user approval. This crosses the boundary between a search request and a financial action, allowing an erroneous or compromised API response to steer the agent toward an unwanted or attacker-controlled payment. The script does not itself transfer funds, and downstream payment safeguards are not shown, but the skill metadata explicitly reinforces automatic recharge.

Content

No source excerpt is available for this finding.

Everyday conversational statements are treated as search triggers without an explicit search request or Douyin context.

Medium
Category
Trigger Abuse
Confidence
94% confidence
Finding

The usage table explicitly maps the conversational observation “最近搞笑视频挺火的” to a search, although it contains neither a search request nor explicit Douyin context. This can cause unintended external API requests, transmitting conversation-derived keywords and consuming account credits. The metadata's automatic Alipay recharge-launch behavior increases the consequences of accidental activation, although this README does not establish that payment can be completed without user authorization.

Content

No source excerpt is available for this finding.

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

Generic terms such as “查询” (query) and “搜索” (search) are designated as activation hooks without requiring Douyin context.

Medium
Category
Trigger Abuse
Confidence
93% confidence
Finding

This instruction treats ordinary search terms and phrases such as “热门文章” (popular articles) as skill invocation hooks. Its example, “查询AI热门文章,” contains no Douyin-specific context, creating an activation rule broader than the Douyin-focused description and potentially routing unrelated searches to this skill.

Content

No source excerpt is available for this finding.

Automatically installs a payment CLI without warning the user about local system changes.

Medium
Category
Output Handling
Confidence
93% confidence
Finding

A missing CLI causes the agent to download and execute an installer without an explicit user-approval step for local software changes. A request to search videos does not inherently authorize software installation, and the automatic payment fallback expands that request into a materially different operation. The document does not establish that payment itself can complete without user involvement.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'npx -y @alipay/agent-payment'.

Medium
Category
MCP Rug Pull
Confidence
97% confidence
Finding

The command executes an automatically accepted, mutable @latest npm release rather than a reviewed, fixed version. A compromised package or publisher account could therefore execute attacker-controlled code with the agent's local permissions; the payment context increases the sensitivity of potentially accessible data. This is a CLI supply-chain issue, not an MCP-server issue as the static finding suggests.

Content

No source excerpt is available for this finding.

Payment-link shortening transmits payment data without a privacy disclosure.

Medium
Category
Output Handling
Confidence
88% confidence
Finding

The workflow directs the agent to send the complete cashier URL, product information, and amount to the payment-link service. The document notes that these URLs contain timestamps and signatures, but does not instruct the agent to disclose this third-party data sharing to the user.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'npx -y @alipay/agent-payment'.

Medium
Category
MCP Rug Pull
Confidence
95% confidence
Finding

The failure-handling text repeats the same unpinned @latest installation command, preserving the supply-chain exposure identified at line 62. This is a duplicate reference to that installation path, not evidence of a second independent execution or an MCP server.

Content

No source excerpt is available for this finding.

A casual observation is treated as a search trigger without requiring search intent or Douyin context.

Low
Category
Trigger Abuse
Confidence
92% confidence
Finding

The usage guide maps a casual statement such as “Funny videos seem hot lately” directly to a search, despite no explicit request or Douyin context. If followed as an activation rule, this could send conversational keywords to the external service and consume API credits without clear user intent. The metadata's automatic recharge flow increases the concern, although this README does not establish that any payment can occur without confirmation.

Content

No source excerpt is available for this finding.