analyzing-on-chain-data

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

• Third-party content exposure detected (high risk: 0.70). The skill explicitly fetches and ingests data from public third‑party APIs (e.g., DeFiLlama at https://api.llama.fi, yields.llama.fi, CoinGecko) as shown in SKILL.md, references/implementation.md and scripts/data_fetcher.py, and that external JSON is parsed and used to calculate metrics and drive rankings/alerts—so untrusted third‑party content is read and can materially influence agent decisions.