AI Native Sdlc Skills — Testing skill for Claude Code

AI-Native SDLC skills

**Unofficial.** This repository repackages Anthropic's **[The AI-Native SDLC playbook](https://claude.com/blog/the-ai-native-sdlc-playbook)** (Louis Claxton, August 2026) as [Claude Code skills](https://code.claude.com/docs/en/skills), so the guidance loads at the moment it applies. The ideas, structure and examples are Anthropic's, not mine. See [NOTICE](NOTICE.md).

The playbook's premise, in one line: **code is no longer the bottleneck — the human-speed steps to the left and right of it are.** Plan, review, test and deploy become the constraint once an agent writes most of the diff, and controls designed around a person reading every line stop matching reality. The answer isn't less process. It's the same control objectives with new enforcement.

Eleven skills carry that into a session.

The artifact chain

idea / ticket / alert
   ↓  intent.md      what is wanted, why, under what constraints   [sdlc-intent]
   ↓  spec.md        requirements + design, concerns flagged       [sdlc-spec]
   ↓  plan.md        files, order, risks, proof                    [sdlc-plan]
   ↓  diff + tests   implementation that verified itself           [sdlc-feedback-loop]
   ↓  PR + findings  agentic review passes, human judges risk      [sdlc-pr-review]
   ↓  production     control bands watch it                        [sdlc-close-loop]
   ↺  breach → new intent.md

Every stage ends by committing something the next stage reads, so the chain of commits *is* the audit trail: who asked for what, what the agent produced, who approved it. Humans stay accountable for every judgment call — what changes is which artifact they read.

The skills

Skill	What it carries
ai-native-sdlc	Router. The artifact chain, which skill fits which situation, adoption order by dependency, and how to read leading vs lagging metrics as a pair
sdlc-intent	Capture an idea, ticket or alert as intent.md — brainstorm-first interview, the tem