Add cognitive on-ramp and project context

[raelovejoy]

Summary

• add a concise, tailored START-HERE.md cognitive on-ramp
• add a structured PROJECT-CONTEXT.md for contributors, maintainers, assistants, and tools
• map canonical sources instead of copying their substance
• separate current reality from ambition, speculation, prototype evidence, and canon
• identify project boundaries, repository structure, working rules, and typed ecosystem connections
• link the short on-ramp to the comprehensive context layer

Why

The project can remain deep without requiring a new or returning person—or an assistant—to reconstruct its purpose and authority model from scattered files. One shared Markdown contract avoids maintaining separate human and AI summaries that could drift apart.

Documentation layers

1. START-HERE.md: smallest useful mental model.
2. PROJECT-CONTEXT.md: comprehensive working briefing and source-of-truth map.
3. Canonical project documents: authoritative detail.
4. Ecosystem map and Wiki: cross-project navigation and relationships.

Maintenance boundary

Review these summaries only when a change materially affects purpose, present status, active frontier, top-level structure, routing, evidence, privacy, authority, or canon. Automation may propose draft changes but must not publish semantic rewrites directly to the default branch.

Existing draft overlap

README.md is also touched by open draft PR #2; review merge order and preserve both changes.

Review focus

1. Is “current reality” strictly accurate?
2. Does the source map point to the actual canonical documents?
3. Are project relationships directional and non-authoritative?
4. Does anything expose private material or overstate evidence/status?