Skip to main content
Omnifence publishes an agent skill that teaches a coding agent how to integrate the moderation API. The skill is agent-agnostic. It installs with the skills CLI, so it works with any coding agent that supports the skills standard, including Claude Code, Cursor, and Codex. The agent finds the places where your codebase creates or accepts content, confirms them with you, and inserts the correct moderation call at each one. Text moderation runs first, and media moderation runs on the result.

​
Install

Install the skill into your project with the skills CLI:
Claude Code can also install the same skill as a plugin, if you prefer to manage it that way:

​
Use

Open your project and ask the agent to integrate:
The skill walks the agent through four stages:
  1. Discovery. The agent scans your codebase for the call sites that produce content. It covers third-party APIs and in-house services alike. It finds in-house platforms by behavioural signals, such as prompt payloads, job queues, and uploads of new media, rather than by SDK names.
  2. Confirmation. Before it writes any code, the agent shows you the list of call sites it found and asks you to confirm, remove, or add entries. Review this list carefully. A missed call site leaves content unmoderated.
  3. Integration. For each confirmed site the agent wires the matching endpoint: POST /api/v1/moderate/text for user-written prompts and AI-character chat turns, /moderate/image, /moderate/video, and /moderate/audio for media.
  4. Async handling. The integration holds content until a pass decision arrives, by webhook or by rate-limit-aware polling, and fails closed on errors.
You need an API key with the moderate:* scopes for the modalities you integrate. The agent reads the key from an environment variable. It never hard-codes the key and it never logs the key.

​
Source

The skill lives at github.com/Omnifence/omnifence-skills.