stack-api

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

• Third-party content exposure detected (medium risk: 0.30). The skill’s runtime uses the 1sat-stack API endpoints (notably /search?q=... and /inscriptions/{id}/content / /content/{outpoint}...) to fetch and ingest outsider-supplied free-text query strings and potentially text content from on-chain inscriptions into the LLM’s prompt context.

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

• Direct money access detected (high risk: 1.00). The skill documents specific blockchain/crypto endpoints (e.g., POST /arcade/tx and /arcade/txs to broadcast transactions, Paymail receive/broadcast endpoints, and UTXO/BSV21 token endpoints and examples). These are explicit facilities for constructing and broadcasting on-chain transactions (i.e., moving crypto funds), so the skill grants direct financial execution capability.