helium-mcp

Pass

Audited by Gen Agent Trust Hub on Apr 16, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill configures a connection to a remote Model Context Protocol (MCP) server at https://heliumtrades.com/mcp to provide its core functionality. This is the intended design for this skill to access the vendor's data services.\n- [PROMPT_INJECTION]: The get_bias_from_url tool retrieves and processes content from third-party article URLs, which creates a surface for indirect prompt injection where hidden instructions on the target site could influence agent behavior.\n
  • Ingestion points: The url parameter in the get_bias_from_url tool defined in SKILL.md.\n
  • Boundary markers: Absent; the skill does not provide specific instructions to the agent to isolate or ignore instructions contained within the fetched content.\n
  • Capability inventory: Network retrieval and text synthesis tools are used to process external content.\n
  • Sanitization: Not specified; external content is analyzed for bias markers without explicit mention of instruction filtering.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 16, 2026, 11:35 AM