Source and attribution policy

Trenchy publishes concise editorial summaries or clearly labelled source-provided descriptions, exact source timestamps, attribution, and a deep link to the original publisher. Full articles and copyrighted media are not republished.

When production storage and scheduling are configured, the automated catalog checks a compile-time allowlist of official Arc publisher pages every two hours. Each successful run can add at most three fresh items. It never uses an LLM to write or expand factual claims.

Automated descriptions are shortened, stripped of markup and unsafe controls, and labelled as publisher-provided metadata. URLs are restricted to credential-free HTTPS on an allowlisted host, redirects are revalidated, and responses have strict timeout and size limits.

Items are deduplicated by stable source identity plus normalized title and description similarity. If fewer than three trustworthy fresh items exist, Trenchy publishes fewer; when no new item exists, it publishes nothing. Fetching, validation, or persistence failures also publish nothing and keep the static verified archive available.