approve-exempt

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.30). The skill monitors and consumes pending security exemptions via harness_list, which reads external request data submitted by users seeking exemptions.

MEDIUM W009: Direct money access capability detected (payment gateways, crypto, banking).

• Direct money access detected (high risk: 0.70). The skill implements direct financial/operational execution (approving security exemptions and elevating their scopes up to account-wide via API calls), but it operates via an MCP server (harness-mcp-v2) with no documented independent access control enforced by the skill itself, relying entirely on the agent's prompts and user interaction.