paperclip-board

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

• Third-party content exposure detected (medium risk: 0.30). In SKILL.md the board-agent startup/monitoring workflow repeatedly reads company issues (including “Board Operations” issue decision-log and task comments) via Paperclip API endpoints like /api/companies/$PAPERCLIP_COMPANY_ID/issues?... and /api/issues/{issueId}/comments, where outsiders can supply free text by posting issues/comments.

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

• Direct money access detected (high risk: 1.00). The skill explicitly includes API fields and example calls that create and modify monetary budgets (e.g., "budgetMonthlyCents" when creating a company and when hiring agents), shows a config revision that changes an agent's budget, and exposes endpoints to update company/agent records and to approve requests. Those examples demonstrate the skill can programmatically update budget values and approve paid items (e.g., approvals for purchases), which constitutes direct financial execution (managing/updating budgets and approval of spend).