The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: Fetched data from several well-known and reputable news organizations.
Evidence: Downloads RSS feeds from BBC (feeds.bbci.co.uk), Reuters (reutersagency.com), NPR (feeds.npr.org), and Al Jazeera (aljazeera.com).
Evidence: Interacts with OpenAI's official API (api.openai.com) for text-to-speech services.
[COMMAND_EXECUTION]: Uses local shell utilities to process fetched news data.
Evidence: Utilizes curl, grep, sed, and head to parse XML data and extract headlines.
[PROMPT_INJECTION]: Vulnerable to indirect prompt injection (Category 8) due to the processing of untrusted external content.
Ingestion points: External RSS feeds from multiple news providers fetched via curl commands in SKILL.md.
Boundary markers: Absent; fetched headlines and descriptions are directly intended for LLM summarization without explicit delimiters.
Capability inventory: Shell command execution (curl, grep, sed), local file writing (/tmp/news.mp3), and network requests to OpenAI API.
Sanitization: Absent; only basic HTML tag removal is performed via sed, which does not prevent instructional content within the news text from influencing the agent.

news-summary