crypto-algo-trader

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

• Third-party content exposure detected (medium risk: 0.30). The skill ingests outsider-authored free text via its sentiment workflow (“Fetch headlines, tweets, Reddit posts”) and runs it through VADER/TextBlob at runtime (scripts/sentiment.py, sentiment signals described in SKILL.md), so an outsider can poison the monitored social sources.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill references and at runtime attempts to load the Amazon Chronos-Tiny model (model name 'amazon/chronos-tiny') which will cause ChronosPipeline.from_pretrained to fetch artifacts from https://huggingface.co/amazon/chronos-tiny and those remote model weights directly influence the agent's predictions/behavior.