Back to skill

Security audit

Doubao API Toolkit

Security checks for vulnerabilities and agentic risk

Overview

This appears to be a legitimate Doubao API toolkit, but it needs Review because its video-analysis command can upload any readable local file path to Volcengine without validation or confirmation.

Install only if you are comfortable sending prompts, image URLs, and chosen local videos to Volcengine ARK using your ARK_API_KEY. Do not let untrusted prompts or automated workflows choose analyze-vid file paths; use only non-sensitive media, preferably from a dedicated folder, and consider adding confirmation plus file type and size checks before use. Generated files are stored under ~/.openclaw/workspace/output.

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 (1)

T09 · Insecure Skill Coding Practices

Warning
Location
scripts/doubao_toolkit.py:199
Finding

Arbitrary Local File Disclosure Through Unvalidated Video Upload

Content
View full analysis
Remediation
View remediation
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Taint TrackingDirect Taint Flow, Variable-Mediated Taint Flow, Credential Exfiltration Chain
  • 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 (15)

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
90% confidence
Finding

The skill documents behavior that requires access to environment variables and external network endpoints, but it does not declare any tool scope such as permissions or allowed-tools. This creates a transparency and policy-enforcement gap: a host may not clearly constrain or inform users about the capabilities needed, increasing the risk of unintended secret access or outbound data transmission.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The documentation encourages users to submit prompts, image URLs, and local video files to an external API and notes that outputs are saved locally, but it does not clearly warn about the privacy and data-handling implications. Users may unknowingly transmit sensitive media or prompts to a third-party service and leave generated artifacts on disk, creating confidentiality and compliance risks.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
80% 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 · scripts/doubao_toolkit.py (reported line 73)May include surrounding context.

python
"size": size
        }
        
        response = requests.post(
            f"{BASE_URL}/images/generations",
            headers=self.headers,
            json=payload,

External Transmission

Medium
Category
Data Exfiltration
Confidence
80% 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 · scripts/doubao_toolkit.py (reported line 125)May include surrounding context.

python
"watermark": False
        }
        
        response = requests.post(
            f"{BASE_URL}/contents/generations/tasks",
            headers=self.headers,
            json=payload,

Tainted flow: 'task_id' from requests.post (line 136, network input) → requests.get (network output)

Medium
Category
Data Flow
Confidence
65% confidence
Finding

Data from a source is assigned to a variable that is later passed to a sink, creating a variable-mediated taint flow.

Content

Scanner excerpt · scripts/doubao_toolkit.py (reported line 165)May include surrounding context.

python
def get_video_status(self, task_id: str) -> Dict[str, Any]:
        """Get video generation status / 获取视频生成状态"""
        response = requests.get(
            f"{BASE_URL}/contents/generations/tasks/{task_id}",
            headers=self.headers,
            timeout=30

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The analyze_image method defaults to a Chinese prompt, and similar Chinese-only defaults appear elsewhere, which nudges output behavior toward a specific language without offering a choice. The file is bilingual overall, but the operational defaults still impose a locale preference rather than making language selection explicit.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
80% 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 · scripts/doubao_toolkit.py (reported line 199)May include surrounding context.

python
"max_tokens": 2000
        }
        
        response = requests.post(
            f"{BASE_URL}/chat/completions",
            headers=self.headers,
            json=payload,

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The analyze_video method sets a Chinese default prompt for content analysis, which establishes a specific language preference by default. Because no locale choice or opt-in is provided, this can conflict with organizational language/locale policy expectations.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The analyze_video function reads an entire local video file, base64-encodes it, and sends it to a remote API, but the user is not explicitly warned at the point of use that local file contents will leave the machine. In a skill/tool context, this can cause unintended disclosure of sensitive recordings, screen captures, or embedded metadata.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
96% confidence
Finding

This request transmits a full local video file, embedded as base64, to a remote chat/completions endpoint. Because the code operates on a local file path and performs the upload without an explicit upload warning or confirmation, it creates a real risk of unintentional disclosure of sensitive local content.

Content

Scanner excerpt · scripts/doubao_toolkit.py (reported line 253)May include surrounding context.

python
"max_tokens": 2000
        }
        
        response = requests.post(
            f"{BASE_URL}/chat/completions",
            headers=self.headers,
            json=payload,

Tainted flow: 'local_path' from requests.post (line 150, network input) → open (file write)

Medium
Category
Data Flow
Confidence
65% confidence
Finding

Data from a source is assigned to a variable that is later passed to a sink, creating a variable-mediated taint flow.

Content

Scanner excerpt · scripts/doubao_toolkit.py (reported line 273)May include surrounding context.

python
local_path = self.output_dir / filename
        
        response = requests.get(url, timeout=120)
        with open(local_path, "wb") as f:
            f.write(response.content)
        
        return local_path

Description-Behavior Mismatch

Low
Category
Not specified by scanner
Confidence
91% confidence
Finding

The manifest description explicitly includes image-to-image support, and the models table also lists an image-editing model. However, the usage section and API reference in this file do not provide any image-to-image command or method, creating a mismatch between claimed capability and the actual exposed behavior described here.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Low
Category
Not specified by scanner
Confidence
74% confidence
Finding

The top-level description presents a broader capability set, including image/video analysis and additional functions, but the concrete interface documented later is narrower and somewhat inconsistent. This is a mild description-to-behavior mismatch because the high-level claims are not cleanly matched by a complete, corresponding documented interface in the same file.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
88% confidence
Finding

The client creates a persistent output directory under the user's home directory and later saves downloaded images and videos there. While the code logs generation progress, it does not clearly disclose in the command help or user-facing messaging that files will be stored locally in this path.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
79% confidence
Finding

The manifest description hard-codes both English and Chinese text in the skill metadata, implying a fixed language/locale presentation rather than offering user choice. Under the stated policy, locale constraints should be optional or explicitly justified as region-specific; no such opt-in or justification appears here.

Content

No source excerpt is available for this finding.