Discord Ban-Recommendation Agent with Human Review Gate

For an offender flagged as ban-worthy, an agent reads their entire ledger, weighs it against your written moderation policy, and produces a structured recommendation: ban, temp-mute, or warn, with the evidence and policy clauses cited. It posts the recommendation to a review queue channel where a moderator approves or rejects — the bot never bans autonomously.

Use it when bans carry real cost (paid members, regulated communities) and every removal must be defensible. The agent does the tedious evidence-gathering; the human keeps the final call.

1. 1A Discord trigger fires when a user is added to the ban-review list (a command or reaction).
2. 2The agent pulls the offender's full history from Postgres and the current policy text from Notion.
3. 3The agent reasons over both and produces a structured verdict with cited evidence and a confidence level.
4. 4A logic step formats the verdict into a review card.
5. 5The bot posts the card to the review-queue channel with approve and reject controls, holding for a human decision.