ClawHub
Back to skill

Security audit

Deepseek WebSearch

Security checks for vulnerabilities and agentic risk

Overview

This is a straightforward web-search skill that sends user search queries to RedFox/Deepseek using a documented API key, with no evidence of hidden persistence, destructive actions, or unrelated data access.

Install only if you are comfortable sending your search terms to RedFox/Deepseek and storing a REDFOX_API_KEY in your environment. Avoid using it for secrets, private documents, credentials, or sensitive personal data unless your organization permits that third-party processing.

SkillSpector

Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (8)

Lp3

Medium
Category
MCP Least Privilege
Confidence
90% confidence
Finding
The skill documentation indicates use of environment variables and outbound network access via an API key and remote search service, but no declared permissions are present. This creates a transparency and governance gap: the host or user may invoke the skill without clear awareness that it reads secrets from the environment and sends user queries to an external service.

Vague Triggers

Medium
Confidence
93% confidence
Finding
The skill instructs users to 'describe your search need in natural language to start,' which creates very broad activation semantics without clear boundaries on when the skill should be invoked. In an agent ecosystem, this can cause over-triggering on generic requests involving search, research, verification, or current events, potentially routing user data or intent to an external web-search capability unexpectedly.

Vague Triggers

Medium
Confidence
95% confidence
Finding
The usage guidance repeats the same vague invocation model—'Simply describe your search need in natural language'—without constraining scope or requiring user confirmation before external lookup. This increases the chance that common-language prompts will activate the skill unintentionally, leading to unnecessary external requests, privacy leakage of prompt contents, and unpredictable tool selection behavior.

Vague Triggers

Medium
Confidence
88% confidence
Finding
The trigger phrases are extremely broad, covering generic terms like '联网搜索', 'AI搜索', and '搜索', which can overlap with ordinary user requests. This can cause the skill to activate unexpectedly and route user prompts or sensitive queries to an external search provider when the user did not specifically intend to use this integration.

Vague Triggers

Low
Confidence
82% confidence
Finding
The usage guidance tells users to describe search needs in natural language but does not define boundaries, exclusions, or safety checks for when the skill should not run. In context, this makes accidental over-invocation more likely and increases the chance that confidential or unnecessary data is transmitted to the external search backend.

Missing User Warnings

Medium
Confidence
95% confidence
Finding
The workflow explicitly sends user-provided search text to external Deepseek API endpoints, but it does not instruct the agent to disclose that data sharing to the user or obtain consent. This creates a privacy and data-handling risk, especially if users include sensitive, personal, or confidential information in their queries expecting only local processing.

Missing User Warnings

Medium
Confidence
85% confidence
Finding
The script sends the user-provided search text to an external third-party API without an explicit runtime notice or consent checkpoint. In an agent-skill context, users may enter sensitive data assuming local processing, so undisclosed transmission can cause privacy leakage or policy violations.

External Transmission

Medium
Category
Data Exfiltration
Content
# ── Step 1: 提交搜索 ──────────────────────────────────────────────
    try:
        submit_resp = requests.post(
            f"{API_BASE}/dsSubmit",
            json={"inquiryText": query},
            headers=headers,
Confidence
82% confidence
Finding
The code transmits user input to an external service, which is inherent to a web-search skill, but the transmission is still security-relevant because it can expose sensitive prompts or personal data to a third party. In this skill context the behavior is expected, yet it remains dangerous if users are not clearly informed and if sensitive data is passed through unredacted.

Static analysis

No suspicious patterns detected.