The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted content from multiple external news platforms and arbitrary URLs, which could contain malicious instructions designed to influence the agent's behavior.
Ingestion points: Data is extracted from external websites via ContentExtractor.extract_with_jina (in scripts/content_extractor.py) and NewsNowTools.fetch_hot_news (in scripts/news_tools.py).
Boundary markers: The instructions do not define clear delimiters or warnings to the agent to ignore instructions embedded within the fetched news content.
Capability inventory: The skill uses requests for network operations and maintains a local SQLite database (data/signal_flux.db) for news items.
Sanitization: No content validation, escaping, or filtering is performed on the extracted news text before it is returned to the agent context.
[EXTERNAL_DOWNLOADS]: The skill connects to several external endpoints to retrieve its core data.
Fetches news aggregation from https://newsnow.busiyi.world.
Retrieves active prediction market summaries from the official Polymarket API at https://gamma-api.polymarket.com.
Uses the Jina Reader service at https://r.jina.ai/ to convert web content into markdown.

alphaear-news