Back to skill

Security audit

元宝搜索Skill

Security checks for vulnerabilities and agentic risk

Overview

This skill provides Tencent web search with disclosed API-key use and no hidden collection or background activity found.

Search terms leave your environment for Tencent and may consume service quota. Keep queries free of secrets and unnecessary personal information, and require confirmation for sensitive searches. Protect the locally stored API key. All bundled files were reviewed; live service behavior and data-retention practices were not verified.

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
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • 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
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Findings (4)

Code capability 'env' detected in scripts/websearch.py but not covered by declared permissions.

High
Category
MCP Least Privilege
Confidence
75% confidence
Finding

The skill uses 'env' capability that is not covered by the declared 'allowed-tools' preapproval. Review the host runtime's actual enforcement policy; runtime permissions are unknown unless separately verified, and this declaration alone does not establish a permission ceiling.

Content

No source excerpt is available for this finding.

Conversation-derived search queries are sent to a third-party service without a privacy warning.

Medium
Category
Output Handling
Confidence
91% confidence
Finding

The skill searches using key questions extracted from the user's conversation through Tencent's Yuanbao search service. Although internet searching is described, the README does not warn that conversation-derived content leaves the agent environment or caution against including sensitive information in queries.

Content

No source excerpt is available for this finding.

Conversation-derived search queries are sent to an external service without a privacy warning.

Medium
Category
Output Handling
Confidence
91% confidence
Finding

The instructions derive external search queries from conversations without specifying sensitive-data exclusion, query minimization, or confirmation before sensitive disclosure. The automatic activation rules make this more than a missing notice: private conversation details could be sent to Tencent even when the user did not explicitly request a search.

Content

No source excerpt is available for this finding.

Broad date and recency triggers can activate web search without a search request.

Medium
Category
Trigger Abuse
Confidence
98% confidence
Finding

The skill requires activation for broad recency and date cues, including a year reference, rather than requiring an actual need or request for external retrieval. This can turn an otherwise local conversation into an external search, potentially disclosing sensitive query terms and incurring unintended service usage.

Content

No source excerpt is available for this finding.