solana-simd

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

• Third-party content exposure detected (medium risk: 0.30). In SKILL.md, the runtime workflow instructs the agent to WebFetch SIMD specs from outsider-facing GitHub raw content (list proposals directory, then fetch specific markdown by filename) and to parse/summarize that fetched text.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill instructs the agent at runtime to fetch and ingest remote specs from https://api.github.com/repos/solana-foundation/solana-improvement-documents/contents/proposals and https://raw.githubusercontent.com/solana-foundation/solana-improvement-documents/main/proposals/{full-filename}, which directly supplies external markdown used to control the agent's responses, so these runtime fetches are a high-confidence external dependency.

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

• Direct money access detected (high risk: 1.00). The skill contains explicit Solana blockchain operations that perform or authorize value transfers: examples showing token transfers (token::transfer), PDA-signed transfers (invoke_signed), creating mints with transfer-fee configuration, submitting and signing transactions via connection.sendTransaction / tx.sign, setting priority fees (ComputeBudgetProgram.setComputeUnitPrice), and closing accounts to reclaim lamports. These are specific crypto/blockchain APIs for moving funds and managing token minting/fees — i.e., direct financial execution capability.