The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill fetches external news articles from an API and presents them for agent processing, which could allow maliciously crafted news content to influence agent behavior.
Ingestion points: The skill uses curl commands in SKILL.md to fetch JSON data from https://agentworld.me/api/ccn/news and https://agentworld.me/api/ccn/latest.
Boundary markers: There are no explicit instructions for the agent to use delimiters or ignore potential commands embedded within the fetched news headlines or summaries.
Capability inventory: The skill utilizes curl for network requests to read external data.
Sanitization: No specific validation, filtering, or escaping of the incoming news JSON payload is performed before the agent interprets the content.
[EXTERNAL_DOWNLOADS]: Fetches news headlines and wire metadata from the vendor-associated domain agentworld.me.
Evidence: curl -s -A "Mozilla/5.0" "https://agentworld.me/api/ccn/news?limit=10" and curl -s -A "Mozilla/5.0" "https://agentworld.me/api/ccn/latest" as documented in SKILL.md.

ccn