The guardrails
Ecosystem tools fail one of two ways: they spam communities until they're banned, or they burn the humans whose names they borrow. Ravenry's rules against both aren't policies — they're enforced in code and tested on every release.
Communities are never auto-posted. Ever.
There is no code path that posts to a community. Drafts are handed to a human who posts from their real account or doesn't. This is the founding invariant and it is not configurable.
Consent is structural
Work is assigned to identities — real people or brand accounts. An identity that hasn't opted in is never assigned anything. Creating an identity for someone else starts opted-out; only they (or an agreed scope) turn it on. Weekly capacity limits how much anyone is asked to do.
People aren't targets
A global 21-day contact cap applies to every person in the map, across all your products. Three interesting posts from one author in a week is one relationship touch, not three emails — same-site suggestions get folded into a single note that references the rest.
Sensitive contexts are hard stops
Signals touching tragedy, controversy, or vulnerable situations are blocked before drafting — regardless of how relevant they are. No draft is ever generated, so none can slip through.
Only public contact information
Enrichment records an email only when the person publishes it themselves. No pattern-guessing, no scraped databases.
Every decision is on the record
Approvals, edits, and rejections land in an append-only decision log — even administrators can't rewrite history. What the system suggested, who decided, and what happened stays auditable.