prompt logo

prompt

Free

Content moderation prompt for AI classification

FreeFree tier
Inputs: textOutputs: text
Type
Open Source

About prompt

This tool provides a prompt designed for content moderation. It instructs an AI to classify user-generated content as ALLOW or BLOCK based on a detailed moderation policy. The policy includes specific categories for blocking (hate speech, threats, CSAM, weapons instructions, spam, PII, illegal content) and conditional allowance with flagging for mature themes, strong opinions, profanity, and sensitive topics. It also includes edge-case guidance for handling sarcasm, reported speech, and fictional content. The output includes a final verdict and, if blocked, a one-line reason for the policy violation.

Key Features

Classifies user-generated content as ALLOW or BLOCK
Includes specific BLOCK categories: hate speech, threats, CSAM, weapons instructions, spam, PII, illegal content
Provides conditional ALLOW with flagging for mature themes, strong opinions, profanity, sensitive topics
Edge-case guidance for sarcasm, reported speech, and creative fiction
Outputs decision with one-line reason for BLOCK

Pros & Cons

Pros
  • Clear, structured moderation policy
  • Handles nuanced cases like sarcasm and reported speech
  • Customizable for different platforms
  • Free and open source
Cons
  • Requires careful prompt engineering to avoid false positives/negatives
  • Relies on underlying LLM's interpretation ability
  • No built-in API or integration; must be used manually or via code
  • May not cover every edge case

Best For

Moderating user comments on websitesFiltering forum posts for policy violationsPre-screening user-generated content before publicationAssisting human moderators in content review

FAQ

How does this prompt classify content?
It instructs the AI to analyze user content against a defined moderation policy and output a decision of ALLOW or BLOCK, with a one-line reason for BLOCK.
What types of content are blocked?
Content containing hate speech, explicit threats, child sexual abuse material, instructions for weapons of mass destruction, spam, personally identifiable information shared without consent, or illegal content.
Can it handle sarcasm?
Yes, the prompt includes edge-case guidance to look for explicit harm signals and distinguish sarcasm from actual harmful content.