Add a Claude Code agent and skill for applying and verifying Varbase AI Safety

Problem/Motivation

Applying this recipe has a step that is easy to miss. The guardrail sets are installed and enabled but attached to no agent or assistant, so nothing is guarded until they are attached to the chatbot, Ask AI and content operations agents by hand. The AI image safety rule on image media is checked by the default vision model, so a site without one has every upload refused or erroring.

An AI coding assistant working on a site that uses the recipe has no guidance for any of this.

Steps to reproduce

• N/A

Proposed resolution

Ship two files in the recipe repository, written from the recipe's own code:

• .claude/skills/varbase-ai-safety/SKILL.md: what the recipe installs, the apply commands, the steps after applying, browser checks and known traps.
• .claude/agents/varbase-ai-safety.md: a thin operator that reads the skill and applies, repairs or verifies the recipe on a DDEV site.

No runtime code or config change. Sites that do not use Claude Code are not affected.

AI-Generated: Yes. Drafted with Claude Code.

Remaining tasks

• ✅ File an issue
• ❌ Addition/Change/Update/Fix
• ❌ Testing to ensure no regression
• ➖ Automated unit/functional testing coverage
• ❌ Developer Documentation support
• ➖ User Guide Documentation support
• ➖ UX/UI designer responsibilities
• ➖ Accessibility and Readability
• ❌ Reviewed by a human
• ❌ Code review by maintainers
• ❌ Full testing and approval
• ❌ Credit contributors
• ❌ Review with the product owner
• ✅ Update Release Notes
• ✅ Release varbase-11.0.x, varbase_ai_safety-1.0.3

User interface changes

• N/A

API changes

• N/A

Data model changes

• N/A

Release notes snippet

• Added a Claude Code agent and skill for applying and verifying Varbase AI Safety.