Enterprise Pattern — Skills

Pattern: Enterprise
Component: skills.md
Version: 1.1 | Updated: 2026-07-16

---

Skill	Domain path	Why required
Responsible AI / risk	09-enterprise-ai/governance/	Tiering and oversight obligations
Least privilege	09-enterprise-ai/security/least-privilege.md	Tool and data scope minimization
OAuth for MCP	09-enterprise-ai/security/oauth-for-mcp.md	Safe tool credentials
Prompt-injection defense	09-enterprise-ai/security/prompt-injection-defense.md	Untrusted content boundaries
Audit logging	09-enterprise-ai/governance/audit-logging.md	Evidence and forensics
Sandboxing	09-enterprise-ai/security/sandboxing.md	Contain tool side effects

Skill activation triggers

1. Risk — every new use case and every material change
2. Least privilege / MCP OAuth — any tool with side effects
3. Injection defense — any untrusted user or retrieved content in context
4. Audit logging — before pilot enablement
5. Sandboxing — code execution or high-impact tools

Verification that skills worked

• Risk: tier matches rubric criteria written in the package
• Privilege: tool allowlist is deny-by-default
• Logging: redaction rules exist for sensitive fields
• Kill switch: drill log shows disable without engineering heroics