How does the AI Safety skill handle moderation?

It applies a dual-layer moderation strategy, filtering inputs before they reach the LLM and validating outputs before they are rendered to the user or applied to the system.

What is the benefit of the structured audit event feature?

It generates a detailed, machine-readable log for every action taken by the AI, ensuring complete transparency and accountability for automated tasks.

Is this skill specific to a certain AI model?

While designed with a focus on Ollama tool-calling features, its safety patterns and moderation gates can be applied to various LLM integration workflows within Claude Code.

Can I require human approval for certain AI actions?

Yes, the skill includes built-in controls to mandate parent or administrator approval for any high-impact changes the AI attempts to execute.

Does this skill support rolling back AI-generated changes?

Yes, it enforces the implementation of a revert path for every mutation, allowing users to safely undo any action performed by the AI tool.

AI Safety & Moderation

Name: AI Safety & Moderation
Author: jakubkrzysztofsikora

byjakubkrzysztofsikora

0•

安全与测试

Implements robust moderation gates, child-safety protocols, and auditable workflows for AI-powered tool-calling features.

The AI Safety skill provides a comprehensive framework for securing LLM interactions, specifically targeting implementations using Ollama tool-calling. It enables developers to enforce input and output moderation, mandate parental or administrative approval for high-impact system changes, and maintain a rigorous audit trail of all AI-driven actions. By integrating structured revert paths and a policy-driven restriction matrix, this skill ensures that AI behavior remains safe, predictable, and compliant with child-safety and enterprise security standards.

主要功能

01Structured audit logging for every AI-triggered action

02Dual-layer input and output moderation gates

03Policy-driven restriction matrix for mode-specific control

040 GitHub stars

05Automated revert paths for undoing AI changes

06Approval workflows for high-impact system mutations

使用场景

01Securing autonomous agents to prevent unauthorized system modifications

02Building child-safe educational applications powered by local LLMs

03Implementing compliance-heavy administrative tools requiring full auditability

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add jakubkrzysztofsikora/choyce-engine ai-safety

For use in Claude.ai and ChatGPT

主要功能

01Structured audit logging for every AI-triggered action

02Dual-layer input and output moderation gates

03Policy-driven restriction matrix for mode-specific control

040 GitHub stars

05Automated revert paths for undoing AI changes

06Approval workflows for high-impact system mutations

使用场景

01Securing autonomous agents to prevent unauthorized system modifications

02Building child-safe educational applications powered by local LLMs

03Implementing compliance-heavy administrative tools requiring full auditability

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add jakubkrzysztofsikora/choyce-engine ai-safety

For use in Claude.ai and ChatGPT