oasis-dev

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

• Third-party content exposure detected (high risk: 0.80). The skill explicitly instructs the agent to fetch external documentation from MCP Context7 (https://context7.com/llmstxt/oasis_io_llms_txt) and to "consult references" (public docs/GitHub links) in its SKILL.md workflow, so it will read and act on third‑party public web content that can influence subsequent actions.

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

• Potentially malicious external URL detected (high risk: 1.00). The skill specifies using MCP Context7 at https://context7.com/llmstxt/oasis_io_llms_txt to fetch latest docs at runtime, which would be injected into the agent context and thus can directly control prompts/responses.

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

• Direct money access detected (high risk: 1.00). The skill is explicitly about interacting with the Oasis blockchain and includes concrete, purpose-built commands and APIs for moving tokens and signing transactions. Examples in the prompt: wallet creation commands, explicit transfer syntax ("oasis account transfer --network testnet --paratime sapphire"), deposit and withdraw commands, guidance on configuring accounts with private keys in Hardhat, and the ROFL appd REST API for key generation and transaction signing. These are specific crypto/blockchain capabilities (wallets, signing, sending transactions) that enable direct financial execution.

MEDIUM W013: Attempt to modify system services in skill instructions.

• Attempt to modify system services in skill instructions detected (high risk: 1.00). The skill explicitly instructs using sudo to move a binary into /usr/local/bin (sudo mv), which tells the agent to perform a privileged system modification and thus can compromise the machine state.