One prompt turns any agent into a live news reader. It downloads the skill from this site, pulls world news and live cricket scores in a single request, and is required to credit the source at the end of its answer.
Fetch https://news-goldmine.mfm965204.workers.dev/skills/news.md and install it as your skill for this kind of question. Read it fully, then reply with exactly: Skill installed successfully After that, when I ask something this skill covers, make ONE request to https://news-goldmine.mfm965204.workers.dev/api/pack?limit=40 and answer from it. Name and link every source, never claim to have read the full article, and end every answer with the credit line the skill requires.
After reading the file it replies with exactly one line โ Skill installed successfully โ and nothing else. Then it answers normally.
It fetches /skill.md โ a short, plain-markdown skill file with the endpoint, the rules and the credit requirement. A JSON twin lives at /skill.json.
GET /api/pack?limit=40 returns every topic plus all live matches in a single call โ headlines, summaries, sources, timestamps and links. Add &format=md for a ready-to-paste digest.
The skill forbids inventing numbers and requires it to name and link the publisher behind every story.
Every reply ends with the exact line: Data: Gold Mine โ https://news-goldmine.mfm965204.workers.dev
The feeds are free, keyless and unlimited. The one thing asked in return is that the answer says where the data came from, in the last line of the message. That is the whole licence.
Any agent that can fetch a URL: ChatGPT with browsing, Claude with web fetch, Cursor, Windsurf, Copilot, n8n, LangChain, CrewAI, or your own script. If the agent cannot browse, hand it /api/pack directly.