prompts.chatprompts.chat
Login
All classifiers
Decision chainModeration2 steps · 6 endpoints

Community post moderation

Sort posts by policy area, then decide how visible they stay based on potential harm.

Scenario paths

Pick a predefined path to highlight it through the chain. No model is called; outcomes are illustrative.

Steps
  1. Community policy category
  2. Harm severity
Endpoints
  • Publish
  • Publish and log
  • Hide as likely spam
  • Collapse and review
  • Hide and escalate
  • Moderator review
What each endpoint means

  • Publish · Automated

    Visible as normal.

  • Publish and log · Automated

    Visible, and logged for moderators' spot checks.

  • Hide as likely spam · Automated

    Hidden until checked. Reversible, and the author can appeal.

  • Collapse and review · Human review

    Collapsed behind a notice while a moderator decides.

  • Hide and escalate · Human review

    Hidden now and sent to the priority moderation queue.

  • Moderator review · Human review

    A moderator decides; the post stays as it is until then.

Steps

  1. Community policy categoryChoiceEntry

    Name the policy area, if any.

    Question, inputs and notes

    Asks:Which part of `community_policy`, if any, does this post most likely break?

    Receives

    • Inputpost
    • Inputsection
    • Inputcommunity_policy
  2. Harm severityScore

    Decide how fast and how visibly to act.

    Question, inputs and notes

    Asks:How much harm could this content cause if it stays visible to the community?

    Receives

About this chain

Goal and input

Handle clear spam with reversible rules, keep civil disagreement visible, and put person-to-person harm in front of moderators quickly.

The post, where it was posted and the community policy.

post
The text of the post or comment.
section
Where it was posted.
community_policy
A short summary of the rules.
How it runs
  • Only reversible actions are automated. Anything that affects an account goes to a moderator.
  • Category and severity read the same post, so both questions can be asked in one request; code applies the severity answer only when the category calls for it.
Sources (3)

Primary documentation this chain builds on.