AI Control Plane
This section defines how agents load context and enforce rules in this repository.
Core principle:
- Canonical engineering and product rules stay in existing docs.
- AI control docs define routing, prioritization, and enforcement.
- No rule source duplication.
Use these pages as the control layer:
- Root Gate
- Specializations
- Context Routing
- Active Rules and Gates
- Tool Mapping
- Approach and Rationale
- Multi-Agent Council
Machine-readable routing source:
configs/ai/context-routing.jsonconfigs/ai/multi-agent-council.json
Operational commands:
pnpm check:policiespnpm metrics:ai-control
Enforced rules
Recurring repository-wide rules are centralized in Root Gate. Other AI docs should reference that page instead of repeating the same rule text.
Naming convention
- Default Markdown filename pattern: lowercase-kebab-case.
- Keep tool-mandated filenames unchanged:
AGENTS.md,PROMPT.md,PROMPTS.md,SKILL.md.