skills/xberg-io/crawlberg/mcp/Gen Agent Trust Hub

mcp

Pass

Audited by Gen Agent Trust Hub on Sep 2, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill documents tools intended to process data from external, untrusted web sources, establishing a surface for indirect prompt injection attacks.\n
  • Ingestion points: Data retrieved via crawl, scrape, and batch_scrape tools listed in SKILL.md allows untrusted content to enter the agent context.\n
  • Boundary markers: The documentation lacks specifications for delimiters or instructions to ignore embedded commands within processed web content.\n
  • Capability inventory: The interact tool allows the agent to execute browser actions that can mutate application state, and the download tool allows fetching external files, both of which could be manipulated by malicious instructions in crawled content.\n
  • Sanitization: There is no evidence of content sanitization, filtering, or validation mentioned in the reference for handling external data.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 2, 2026, 06:27 PM
Security Audit — agent-trust-hub — mcp