ClawHub
Back to skill

Security audit

Chatbot

Security checks for vulnerabilities and agentic risk

Overview

This is a straightforward SenseAudio voice chatbot guide that uses external SenseAudio APIs as expected, with no hidden install, persistence, or destructive behavior found.

Before installing or using this skill, treat SenseAudio API keys, session tokens, room IDs, and conversation IDs as sensitive. Use server-side mediation where possible, avoid logging tokens, and make sure users understand that voice/chat sessions may be processed by SenseAudio services.

A.I.G

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

SkillSpector

Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Findings (7)

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill prominently encourages building voice chatbots but does not clearly disclose that conversation/session metadata and credentials such as room_id, conv_id, and tokens are transmitted to an external third-party service. In a voice assistant context, users may also send sensitive spoken content, so the missing warning undermines informed consent and can lead to unintentional data exposure.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
90% confidence
Finding

The skill hardcodes use of an external SenseAudio API endpoint to retrieve available agents, which means API credentials and request metadata are sent to a third-party service. This is consistent with the skill's purpose, but without prominent disclosure it creates a real privacy/transparency issue for users and deployers.

Content

Scanner excerpt · SKILL.md (reported line 33)May include surrounding context.

md
import requests

def get_available_agents(page=1, size=10):
    url = "https://api.senseaudio.cn/v1/realtime/agents"
    headers = {"Authorization": f"Bearer {API_KEY}"}
    params = {"page": page, "size": size}

External Transmission

Medium
Category
Data Exfiltration
Confidence
91% confidence
Finding

This endpoint creates dialogue sessions with a third-party provider and returns session credentials including conv_id, room_id, and token. In the context of a voice chatbot skill, these values are sensitive and can enable session access if mishandled, so the external transmission and credential issuance should be explicitly called out.

Content

Scanner excerpt · SKILL.md (reported line 58)May include surrounding context.

python
def create_dialogue_session(agent_id, new_dialogue=True, conv_id=None):
    url = "https://api.senseaudio.cn/v1/realtime/invoke"
    headers = {
        "Authorization": f"Bearer {API_KEY}",
        "Content-Type": "application/json"

External Transmission

Medium
Category
Data Exfiltration
Confidence
87% confidence
Finding

This code sends agent_id and potentially conv_id to SenseAudio's external invoke endpoint to create or resume a session, which is expected functionality but still constitutes third-party data transmission. Because the skill is for real-time voice conversations, this call can establish sessions tied to user interactions and should be treated as privacy-sensitive rather than inherently malicious.

Content

Scanner excerpt · SKILL.md (reported line 72)May include surrounding context.

md
if not new_dialogue and conv_id:
        payload["conv_id"] = conv_id

    response = requests.post(url, headers=headers, json=payload)
    return response.json()

# Returns:

External Transmission

Medium
Category
Data Exfiltration
Confidence
84% confidence
Finding

The status check sends room_id and authentication to a third-party endpoint, exposing session metadata externally. This is expected for hosted session management, but it remains a genuine data-sharing behavior that should be disclosed and protected.

Content

Scanner excerpt · SKILL.md (reported line 149)May include surrounding context.

md
def check_status(self):
        """Check if agent is running"""
        url = "https://api.senseaudio.cn/v1/realtime/status"
        headers = {"Authorization": f"Bearer {self.api_key}"}
        params = {"room_id": self.room_id}

External Transmission

Medium
Category
Data Exfiltration
Confidence
84% confidence
Finding

The leave endpoint posts room_id and authorization to an external service to end a session. Although this is normal operational behavior, the session identifier and credentials are sensitive enough that undisclosed or careless use could expose session-management capabilities.

Content

Scanner excerpt · SKILL.md (reported line 158)May include surrounding context.

md
def end_conversation(self):
        """Stop the agent"""
        url = "https://api.senseaudio.cn/v1/realtime/leave"
        headers = {
            "Authorization": f"Bearer {self.api_key}",
            "Content-Type": "application/json"

External Transmission

Medium
Category
Data Exfiltration
Confidence
82% confidence
Finding

This request sends room_id to an external endpoint to terminate a live session. While lower risk than sending conversation content, it still transmits session metadata to a third party and should be disclosed, especially because session identifiers may be sensitive in multi-user systems.

Content

Scanner excerpt · SKILL.md (reported line 165)May include surrounding context.

} payload = {"room_id": self.room_id}

text
    response = requests.post(url, headers=headers, json=payload)
    self.session_active = False
    return response.json()
text

Static analysis

No suspicious patterns detected.