AI-Native SDLC

The SDLC harness for AI-native software delivery

Runtime harnesses govern agent execution.
AI Native SDLC governs software delivery.

AI agents need runtime harnesses to coordinate reasoning, tools, and execution. Software delivery requires planning, specifications, ticket hierarchies, worktrees, reviews, QA, deployments, handoffs, continuity, and observability.

SDLC Harness governs software delivery — the engineering process itself, not a single task.

See Execution Architecture GitHub

npx @ai-native-sdlc/cli install

Three layers: Runtime Harness, AI Native SDLC, Application.

The agent completed a task. The software lifecycle remains unmanaged.

AI execution survives the task.
Software delivery must survive the project.

The operational layer between execution and delivery

Runtime Harness vs SDLC Harness — one governs the agent loop; the other governs delivery operations.

The runtime governs execution. The SDLC Harness governs delivery operations.

Materialized software delivery infrastructure

Hover a folder — purpose, example artifact, and workflow for that harness surface.

• .sdlc/
  • workflows/
  • playbooks/
  • skills/
  • commands/
  • gates/
  • traces/
  • context/
  • manifests/
  • sdlc.yaml

workflows/

Workflow

goal-flow.yaml

Artifact

Composite DSL — 9 nodes

Execution path

/sdlc:goal → runner step loop

From intent to governed delivery

Intent → Specification → Tickets → Worktrees — then review, evidence, deploy, handoff, and trace.

Input Build SaaS onboarding with email verification

1. Intent User prompt → /sdlc:goal
2. Specification .sdlc/specs/JAMBU-122.md
3. Ticket Hierarchy 12 Plane epic + children
4. Worktrees 12 ../worktrees/JAMBU-*
5. Implementation app/src/components/**
6. Review Reviewer gate PASS
7. QA Artifacts 47 .sdlc/evidence/run-id/
8. Pull Requests 12 gh pr create × 12
9. Deployment 1 Cloudflare Pages URL
10. Handoff 1 .sdlc/handoffs/LATEST.md
11. Execution Trace .sdlc/trace/events.jsonl

Delivery artifacts, not chat output

Select an artifact — realistic preview from a governed run, not placeholder copy.

• SpecificationVertical implementation plan with entry point, gates, and verification.
• Ticket HierarchyPlane epic decomposed into parallel vertical slices.
• WorktreesIsolated git worktrees per ticket — no direct main commits.
• QA EvidencePlaywright E2E, traces, and screenshots under .sdlc/evidence/.
• Pull RequestsOne PR per slice with harness_run_id in body.
• Deployment RecordCI deploy workflow with production URL in manifest.
• Handoff PackageSession continuity for the next agent.
• Execution TraceCognitive trace stream and workflow node timeline.

Specification

Vertical implementation plan with entry point, gates, and verification.

## Technical Notes
Entry: /sdlc:goal → goal-flow.yaml
Transform: classify → decomposition → plane-hierarchy
Output: registered worktree + manifest.yaml
Verify: playwright test e2e/smoke.spec.ts

Install and execute

Run install, init, then goal — repository tree, artifacts, and execution graph materialize live.

ai-native-sdlc

▌

spec.md manifest.yaml evidence/ trace.jsonl

0 Tickets

0 Worktrees

0 PRs

0 QA artifacts

Execution graph: idle