The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from external news RSS feeds.\n
Ingestion points: scripts/news_sentiment.py fetches headlines and summaries from Cointelegraph, CoinDesk, and Decrypt RSS feeds.\n
Boundary markers: Untrusted news content is presented to the agent without explicit delimiters or warnings to ignore embedded instructions.\n
Capability inventory: The skill uses Bash to run analysis scripts. While the processing is keyword-based, the resulting headlines are included in the agent's context, providing a path for external content to influence agent decisions.\n
Sanitization: Content is filtered for HTML tags using regex, but no natural language sanitization is performed.\n- [EXTERNAL_DOWNLOADS]: The skill fetches data from established cryptocurrency market APIs and news providers.\n
Evidence: Fetches data from api.alternative.me (Fear & Greed Index), api.coingecko.com (Price Momentum), and several industry RSS feeds.\n- [DYNAMIC_EXECUTION]: The skill uses runtime path modification to load external code modules.\n
Evidence: scripts/news_sentiment.py dynamically updates sys.path to import NewsAggregator from a sibling skill directory (crypto-news-aggregator) if it is present on the filesystem.

analyzing-market-sentiment