helix-mcp-revenue-enablement
Pass
Audited by Gen Agent Trust Hub on Oct 25, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDATA_EXFILTRATION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill's operating principles (Rules 2 and 4 in
references/operating-principles.md) instruct the agent to verify user-supplied claims, URLs, and competitor data usingweb_searchandweb_fetch. This creates a surface where malicious content from external websites could influence agent behavior. - Ingestion points: User-provided URLs and competitor names processed via
web_fetchandweb_search(SKILL.md, references/operating-principles.md). - Boundary markers: Uses
Assumption:prefixes to label unverified data in output and includes a "No-Harmful-Output" rule to mitigate incorrect assertions. - Capability inventory: Access to
web_fetch,web_search, and 12 deal-strategy MCP tools (SKILL.md). - Sanitization: No specific escaping or sanitization of fetched web content is detailed beyond the verification logic.
- [EXTERNAL_DOWNLOADS]: The skill connects to a hosted MCP server at
https://revenue-enablement.gtmhelix.com/mcpand references a vendor-specific npm package@shashwatgtmalpha/revenue-enablement-mcpfor local registration. - [COMMAND_EXECUTION]: Provides CLI registration commands for adding the MCP server to Claude Code using
claude mcp add, which involves executing shell commands with remote URLs. - [DATA_EXFILTRATION]: The skill is designed to process B2B deal strategy and ROI data through a hosted MCP server at
https://revenue-enablement.gtmhelix.com/mcp. Users should be aware that sensitive deal information and business metrics provided to the tools will be transmitted to this external endpoint.
Audit Metadata