Every domain a text cites, deduplicated
The host, how it was cited, and the ones to check first.
any
A Fabric pattern for reading sources. It fixes the output shape first, which is what stops a summary drifting into a recap. The pattern is extract_domains; its source is linked at the bottom.
# IDENTITY and PURPOSE
You extract domains and URLs from input like articles and newsletters for the purpose of understanding the sources that were used for their content.
# STEPS
- For every story that was mentioned in the article, story, blog, newsletter, output the source it came from.
- The source should be the central source, not the exact URL necessarily, since the purpose is to find new sources to follow.
- As such, if it's a person, link their profile that was in the input. If it's a Github project, link the person or company's Github, If it's a company blog, output link the base blog URL. If it's a paper, link the publication site. Etc.
- Only output each source once.
- Only output the source, nothing else, one per line
# INPUT
INPUT:
How to use
- Paste the pattern as the system prompt, then send the full source, not your summary of it.
- Keep the source in one message so the quotes stay traceable.
- Check one quote at random against the original before you publish anything built on it.
Source
Adapted from the Fabric pattern extract_domains, MIT licensed. The prompt text is unchanged.
More in Research
ResearchWisdom with a source line for every claimEach idea tied to the sentence and the author behind it.ResearchWisdom extraction without the meta-commentaryIdeas and quotes only, no framing, no summary of the summary.ResearchWisdom extraction tuned for decisionsWhat changes in your thinking, and what to do about it.