Back to skill

Security audit

Claw Chat Hub

Security checks for vulnerabilities and agentic risk

Overview

This is a coherent chat client, but it needs review because it sends agent messages and history over WebSocket while providing weak or misleading security controls.

Review this before installing in any production or shared environment. Use only trusted Hub endpoints, prefer WSS, do not send secrets or sensitive personal data through it, and do not rely on the api_key parameter for authentication unless the client and Hub are fixed to enforce it.

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 (3)

T09 · Insecure Skill Coding Practices

Error
Location
chat_client.py:14
Finding

API Key Is Ignored and Agent Identity Is Asserted Without Authentication

Content
View full analysis
bool: """连接到 Hub""" try: self._ws = await websockets.connect( self.hub_url, extra_headers={"X-Agent-Id": self.agent_id} ) ``` ### Technical Analysis The public constructor accepts an `api_key`, but the value is only stored and is never transmitted or otherwise used during connection establishment. The WebSocket connection supplies only an `X-Agent-Id` header whose value is controlled entirely by the caller. Consequently, this client provides no cryptographic or credential-based proof that the connecting party owns the asserted agent identity. The ultimate exploitability depends on the Hub implementation: a Hub that independently authenticates the connection may mitigate this condition, but the audited client does not support or invoke that protection. ### Attack Path 1. An attacker learns or guesses the identifier of another agent. 2. The attacker constructs `ChatClient` with the victim's identifier as `agent_id`. 3. `connect()` sends that untrusted identifier in the `X-Agent-Id` header without using the configured API key. 4. If the Hub trusts this header without independent authentication, it registers the attacker as the victim. 5. The attacker can then send messages or request chat and channel operations under the vic ...[truncated 449 chars]
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
chat_client.py:16
Finding

Plaintext WebSocket Transport Is the Documented Default

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
chat_client.py:133
Finding

Chat Acceptance and Rejection Responses Are Not Correlated With Their Request

Content
View full analysis
Remediation
View remediation
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 (5)

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The skill explicitly supports real-time messaging and message history retrieval, which implies transmission and storage of potentially sensitive user or agent data, yet the documentation provides no warning about data handling, retention, access control, or privacy considerations. In a communications skill, this omission can lead integrators to send confidential content without implementing consent, minimization, or retention safeguards.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

User-facing docstrings and runtime messages in this file are written in Chinese, including connection/error output and API descriptions, with no option for user language selection. Per the policy, hard-coding a specific language without user opt-in is a natural-language policy violation unless the locale restriction is explicitly justified.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

This code file sends chat content and metadata over the network via WebSocket in send_message, and similar network transmission also occurs in request_chat and get_history. Although network I/O is core to a chat client, the file provides no user-facing disclosure about transmitting message content, agent identifiers, or chat-history queries, which is safety-relevant for privacy-sensitive use.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
80% confidence
Finding

The natural-language instructions and description are entirely in Chinese, which effectively imposes a language choice on users without offering an alternative or opt-in. This can violate language/locale policy when no justification or user selection is provided.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
88% confidence
Finding

The only natural-language description in this file is written in Chinese ("智能体间实时通讯客户端"), which may indicate the skill is presented in a fixed language without any opt-in or locale selection. The file does not provide any indication that users can choose a language or that the locale restriction is intentionally scoped.

Content

No source excerpt is available for this finding.