outline-wiki-search

Fail

Audited by Gen Agent Trust Hub on Jul 3, 2026

Risk Level: HIGHCREDENTIALS_UNSAFECOMMAND_EXECUTIONDATA_EXFILTRATION
Full Analysis
  • [CREDENTIALS_UNSAFE]: The skill explicitly instructs the agent to access and read sensitive configuration files such as ~/.claude.json and .mcp.json to extract authentication tokens (Authorization: Bearer <token>). This is a direct violation of secure credential management practices.
  • [COMMAND_EXECUTION]: The instructions direct the agent to execute shell commands (curl) to interact with a REST API. This bypasses the brokered MCP (Model Context Protocol) interface, which is designed to provide a secure and audited way for agents to interact with external services.
  • [DATA_EXFILTRATION]: The skill describes a pattern where sensitive data (authentication tokens harvested from local files) is sent over the network via curl. While the intended destination is the user's Outline Wiki instance, this logic creates a high-risk data exfiltration vector where credentials could be diverted to an unauthorized endpoint if the <base> URL is manipulated or incorrectly resolved.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Jul 3, 2026, 06:05 PM
Security Audit — agent-trust-hub — outline-wiki-search