freeagent

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

• Third-party content exposure detected (medium risk: 0.65). The skill’s runtime workflow uses ClawLink to call FreeAgent and retrieve/submit data, and those FreeAgent responses (e.g., company details, invoices/estimates, transaction lists) are outsider-authored free-form text from a third-party SaaS that the agent ingests into its LLM context.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 1.00). The skill instructs the agent to run "npx @useclawlink/cli" which fetches and executes remote code from the npm package/registry (see https://www.npmjs.com/package/@useclawlink/cli), so an external URL is used at runtime to execute remote code.

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

• Direct money access detected (high risk: 1.00). The skill exposes explicit accounting/financial actions (create invoice, create estimate, create bill with attachment, create bank account, create a timeslip, list accounting transactions, cashflow and P&L summaries). These are specific finance-oriented operations that allow the agent to create financial records and manage accounting entities via FreeAgent through ClawLink (which holds OAuth credentials). Because the tool is specifically designed for financial/accounting operations (not a generic browser or HTTP tool) and includes endpoints that create invoices/bills and bank-account records, it constitutes direct financial execution authority.