strand-fact-check-cite

Programmatic blogs lose to fabrication. This skill makes every factual claim in a Strand post traceable to a real source, which is both an ethics floor and a ranking advantage — Google's E-E-A-T and LLM search engines both up-weight cited, verifiable content.

Process

Extract claims. List every checkable assertion: numbers, dates, named events, quotes, prices, legal/regulatory statements, study findings. Opinion and analysis don't need sourcing; facts do.
Verify against primary sources. Use web search/fetch. Prefer the origin — government sites, company filings/blogs, peer-reviewed papers, official statistics — over aggregators. If two reputable sources conflict, say so in the text and cite both.
Cite inline. Attach each verified claim to a source at the point it's made (a linked reference or footnote, per the theme's citation component). Reword source material — never paste source sentences verbatim.
Populate sources[]. Every inline citation maps to a sources[] entry in frontmatter (title, url, optional publisher). The renderer emits these as visible references and as citation/isBasedOn JSON-LD.
Flag the unverifiable. If a claim can't be sourced, either cut it, soften it to attributed reporting ("according to …"), or mark it for human review. Do not publish an unsourced hard fact as established truth.

Rules

Never invent a source, URL, statistic, date, or quote. A plausible-looking fake citation is the worst possible failure for this project.
A live URL you actually retrieved beats a half-remembered citation every time — fetch it.
Quotes are last resort and kept short; paraphrase and attribute instead.
For NewsArticle posts, sourcing is mandatory before strand-publish runs.
Prefer dated sources for time-sensitive claims, and note "as of " when a figure can change.