flare-fassets
Pass
Audited by Gen Agent Trust Hub on Oct 16, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill interacts with external, untrusted data sources including blockchain RPC nodes and Flare Data Connector (FDC) verifiers, which represents an indirect prompt injection surface.
- Ingestion points: Scripts such as
scripts/execute-minting.tsandscripts/get-redemption-queue.tsingest data directly from blockchain state and external verifier APIs. - Boundary markers: The
SKILL.mdfile explicitly identifies external payloads as untrusted and provides instructions to decode them only according to fixed binary formats and contract ABIs, warning against treating them as natural language input. - Capability inventory: The skill includes scripts for signing and broadcasting on-chain transactions, enabling state changes on the Flare and XRP Ledger networks.
- Sanitization: Implementation scripts include input validation such as
isAddress()checks for blockchain addresses. - [COMMAND_EXECUTION]: The skill contains multiple TypeScript scripts (e.g.,
scripts/direct-mint-fxrp.ts,scripts/xrp-payment.ts) intended for execution by the user. These scripts facilitate blockchain interactions and require the use of private keys or seeds supplied via environment variables to sign transactions.
Audit Metadata