A note in this repository, Claude/ai-native-sdlc-playbook/README.md, summarizing The AI-Native SDLC playbook by Louis Claxton, published by Anthropic on 2026-08-21.1 The note points out that it is an Anthropic-authored model built around Claude products, so its product choices are not vendor-neutral evidence.1
Takeaways
- Each delivery stage leaves a committed artifact (
intent.md,spec.md,plan.md, results, PR, incident record) and production evidence loops back as new intent.1 See AI-native SDLC. - Autonomy grows in steps, with the production gate kept human.1 See Risk-based autonomy.
- Deterministic checks enforce; agents judge and diagnose.1 See Least-privilege tool access.