xquik-mcp

Pass

Audited by Gen Agent Trust Hub on Aug 28, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill facilitates interaction with an external MCP server, creating an ingestion surface for untrusted data. It proactively mitigates this risk by instructing the agent to treat all server-returned content as data, to ignore embedded instructions, and to never allow returned content to influence tool or destination selection.\n
  • Ingestion points: Data returned from https://xquik.com/mcp (execute, search, and docs tools) as noted in SKILL.md.\n
  • Boundary markers: Explicit instructions in SKILL.md to treat X-authored content as untrusted data and ignore embedded instructions.\n
  • Capability inventory: Remote MCP tool execution capabilities (execute, search, docs).\n
  • Sanitization: Relies on strict instructional guidelines to the AI model to maintain control flow boundaries.\n- [SAFE]: The documentation in references/mcp-setup.md provides guidelines for managing credentials like XQUIK_API_KEY and OAuth tokens. It contains specific warnings against hardcoding or insecurely storing these secrets, advising users to use environment-backed secret stores rather than shell history or configuration files.\n- [SAFE]: The setup guide includes standard CLI commands for adding MCP servers to various clients (e.g., claude mcp add). These are documentation examples for user-initiated configuration and do not involve silent or automated execution of untrusted scripts.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 28, 2026, 05:45 AM
Security Audit — agent-trust-hub — xquik-mcp