The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses bash and curl to fetch JSON data from a remote API.\n
Evidence: curl -sS -H "User-Agent: ai-daily-brief/1.0.0 (+https://aihot.virxact.com/aihot-skill/)" "https://aihot.virxact.com/api/v1/items?mode=selected&window=24h&limit=15" in SKILL.md.\n- [EXTERNAL_DOWNLOADS]: The skill makes network requests to aihot.virxact.com, a domain not present on the standard whitelist.\n
Evidence: The execution steps define a curl request to https://aihot.virxact.com/api/v1/items.\n- [PROMPT_INJECTION]: The skill processes untrusted external data from two sources, creating an indirect prompt injection surface.\n
Ingestion points: Data is ingested from aihot.virxact.com/api/v1/* (JSON) and github.com/trending (HTML).\n
Boundary markers: The instructions explicitly tell the agent to treat the returned content as "untrusted evidence" and to remind users to check the original links.\n
Capability inventory: The skill uses bash + curl for API calls and WebFetch for web scraping.\n
Sanitization: The agent is instructed to only use the returned content for summarization and is explicitly forbidden from using its training memory to substitute for real-time data.

ai-daily-brief