AI Native Engineering System

AI Native Engineering System

ClawAI treats coding-agent context as a governed engineering subsystem.

Layers

• Root policy / routers — CLAUDE.md, AGENTS.md, CODEX.md, cursor.md, model-specific routers.
• Rules — mandatory constraints.
• Skills — operational runbooks.
• Reviewer agents — specialist review charters.
• Context — stable architecture/navigation maps.
• Memory — durable dated lessons explaining why rules exist.
• Generated manifests — current code/config facts.
• Task packs — compact domain contexts.
• Portable work skills — cross-agent foundations, domain skills and quality gates.
• Knowledge tooling — deterministic classification, manifest generation, verification, benchmarking and task context.

The resolver is deterministic rather than a free-form AI guess:

npm run knowledge:context -- --task="<task>"

See .ai/BOOTSTRAP.md, CLAUDE.md, rules/README.md, skills/README.md, and tools/knowledge.