1. Start with the finished-item care label
Fiber identity is not enough to determine safe care. Dyes, finishes, blends, trims and construction matter, so the permanent care instruction on the actual item is treated as the product-specific authority.
2. Prefer primary and high-authority sources
The source registry records regulators, recognized textile-care organizations, technical references and established cleaning guidance. Source records are linked to content and to specific structured claims where practical.
3. Separate facts from presentation
Fabric risks, stain stages and stain × fabric rules live as structured data. Editorial copy explains those records; it is not the runtime source of truth for the Stain Removal Solver.
4. Fail closed when a combination is not reviewed
The Stain Removal Solver returns a treatment only when the exact stain × fabric combination has an approved, supported rule above the confidence threshold. Unsupported combinations are rejected rather than filled with generated generic advice.
5. Review freshness and duplication
Pages carry fact-check and next-review dates. A fingerprint and near-duplicate audit runs during updates, and reviewed internal links are stored as a separate graph.
6. Require site-level publication gates
A page must pass trust, structured-data, internal-link, duplicate and site-level publication checks before it can be included in a public search sitemap. Page-level settings cannot bypass those release gates.