Verification & Quality Library

Patterns for ensuring AI-generated work is correct, complete, and verified.

When to Use

• Preventing incomplete or hallucinatory work
• Implementing human checkpoints in AI workflows
• Catching common AI hallucinations before code review
• Enforcing proof-of-functionality requirements

References

expected-output-protocol.md

Requires declaring expected outputs BEFORE implementation, then comparing actual results. Creates immediate feedback loops.

human-review-gates.md

Mandatory human checkpoints in AI-assisted workflows. Prevents agent-validates-agent anti-patterns. Based on Capital One's "Writing is Reviewing" principle.