apex-golden-principles

Golden Principles

These principles govern how every agent operates in this repository. They are adapted from the Harness Engineering philosophy for agent-driven infrastructure development.

Rules

The canonical rules and preservation map live in Principle Details. This index is a summary, not a competing policy source.

1. Repository Is the System of Record — all context lives in-repo
2. Map, Not Manual — instructions point to deeper sources
3. Enforce Invariants, Not Implementations — set boundaries, allow autonomous expression
4. Parse at Boundaries — validate inputs and outputs at module edges
5. AVM-First, Security Baseline Always — prefer Azure Verified Modules + non-negotiable security baseline
6. Golden Path Pattern — prefer shared utilities over hand-rolled helpers
7. Human Taste Gets Encoded — turn reviewed feedback into durable guidance
8. Context Is Scarce — reuse current context and recover missing required evidence
9. Progressive Disclosure — load detail when needed without hiding essential rules
10. Mechanical Enforcement Over Documentation — validate deterministic invariants
11. Composable Workflows — bounded steps with explicit ownership and prerequisites
12. Human Approval at Critical Gates — approvals remain evidence-bound and human-controlled
13. Adversarial Review — independent review follows the current graph and role contracts
14. Continuous Lessons — capture process observations and feed them back into the system

Steps

Applying the principles to a new agent or skill:

1. Read the canonical principles before applying this decision framework
2. For each design decision, ask which principles apply (typically 2–3 will dominate)
3. Run the per-principle test listed in Principle Details
4. Where a principle conflicts with an implementation choice, change the implementation — principles are non-negotiable
5. Document proposed deviations in an ADR; an ADR does not authorize relaxing safety or approval gates