[1.12] Add AI-SECURITY.md documentation

[vbonk]

What needs to be done

Dedicated document on AI-assisted development security: prompt injection, attack vectors (issue/PR injection, fork attacks, dependency confusion), how template protects you (CODEOWNERS, branch protection, PR scanner, hooks), best practices, GitHub Actions injection prevention.

Implementation

Include Mermaid diagram showing defense layers. Cross-link from README, CLAUDE.md, AGENTS.md, all AI configs.

Acceptance Criteria

• docs/AI-SECURITY.md exists
• Covers all attack vectors and defenses
• Mermaid defense layer diagram
• Cross-linked from all AI config files

[vbonk]

Completed in Phase 1 PR #87.