Process

Every non-trivial change follows intent -> spec -> plan -> test -> code -> review, in that order. A stage is finished only when its artifact exists on disk, so the work survives /clear, context compaction and a switch of agent.

The rules themselves live in one place: the installed AGENTS.md (in this repository, sdlc_kit/template/AGENTS.md). This page is a map; if it and AGENTS.md ever disagree, AGENTS.md wins.

#	Stage	Artifact	Skill
1	Intent	intent.md	sdlc-plan
2	Spec	spec.md	sdlc-plan
3	Plan	plan.md	sdlc-plan
4	Test	tests	sdlc-implement
5	Code	source	sdlc-implement
6	Review	REVIEW.md	sdlc-review

Two more skills sit around the stages: sdlc finds the current stage (start here, also after /clear), and sdlc-release runs the full gate, finishes the changelog and drafts the PR description. Committing, pushing and publishing happen only when the operator asks.

Start with /sdlc in Claude Code, or say "use the sdlc skill" in another harness.

Reviewing with fresh context

The sdlc-review skill tells the agent to hand only the artifacts and the diff to a fresh subagent or session and to treat its report as data. Claude Code: the Agent tool. Others: a new session or their subagent feature.

Red first

A new test must be seen failing for the right reason before the code that fixes it. The runner proves it with --red; see The gate.