Add compounding systems engineering guidance

[jremick]

Summary

• reposition the guide from prompt abstraction alone toward verified, compounding systems engineering
• add dedicated guides for the three-axis maturity model, graph and ontology-engineered harnesses, and governed improvement loops
• add synthetic templates for system design, graph contracts, and promotion checks
• integrate system-level verification, capability promotion, and orchestration boundaries into the existing guidance
• ground the new material in current official OpenAI harness, Codex, and evaluation sources

Security and public-safety boundary

• keep examples synthetic and prohibit one-to-one publication of private graphs, ontologies, traces, fixtures, state, or configuration
• distinguish observable harness traces from private model reasoning
• require least privilege, separate candidate and promoted state, provenance, independent checks, explicit promotion authority, and rollback
• add reviewed official OpenAI hosts to the external-link allowlist with regression coverage

Validation

• python3 -m unittest discover -s tests -v
• python3 scripts/validate_content.py
• git diff --check
• current official source pages opened successfully on 2026-08-20