A generic, governed AI delivery kit for building, rescuing, reviewing, testing, securing, and shipping software systems from a project-specific spec.
The kit is intentionally generic. It is designed to start building any system from an authoritative spec placed under specs/.
This repository is an operating model for AI-assisted software delivery. It combines:
- lifecycle commands
- reusable delivery skills
- governance rules
- an explicit orchestration layer
- traceability and evidence patterns
- readiness scoring
- rescue workflows
- NFR, architecture, DevOps, AppSec, and PT support
It helps teams move from requirements to implementation with stronger discipline, reviewability, and release readiness.
This repository is not:
- a generated business application
- a domain-specific product template
- a single-stack framework implementation
- a substitute for product thinking, engineering review, or testing judgment
Project-specific requirements belong in specs/, not in the core kit.
Added the next reliability wave on top of V7.3:
- stronger rescue mode
- better traceability automation
- test strategy packs
Extended the generic kit with:
- NFR packs under
packs/ - architecture decision packs under
adr-packs/ - DevOps and deployment packs under
devops-packs/ - supporting skills, templates, examples, and documentation
Integrated Application Security (AppSec) and Penetration Testing (PT) into the governed lifecycle.
Strengthened implementation discipline with:
- assumption and ambiguity control
- simplicity and complexity control
- surgical change control for rescue and maintenance work
- verification attached to each major step or slice
Consolidates the kit with an explicit orchestration layer:
- upgraded
ORCHESTRATOR.mdas the human-facing operating guide - added
skills/orchestration-and-lifecycle-control/as the AI-facing orchestration workflow - updated governance and docs so orchestration is a first-class part of the operating model
The orchestration layer exists to make sure the right parts of the kit are activated in the right order.
Use:
ORCHESTRATOR.md
This file helps operators decide:
- whether the project is new, weak-spec, existing, rescue-stage, or release-stage
- which lifecycle entry point to use
- which supporting skills should be activated
- what artifacts should exist before moving forward
Use:
skills/orchestration-and-lifecycle-control/
This skill helps the AI:
- classify project state
- choose the right command
- activate the right skills and packs
- define minimum artifact expectations
- block invalid lifecycle jumps
- Add your project spec under
specs/. - Read
ORCHESTRATOR.md. - Follow
AGENTS.md. - Start with the correct lifecycle entry:
commands/START-PROJECT.mdfor new systemscommands/SPEC.mdfor weak specscommands/RESCUE-PROJECT.mdfor existing or incomplete systems
- Continue through the lifecycle as appropriate:
commands/PLAN.mdcommands/BUILD.mdcommands/TEST.mdcommands/REVIEW.mdcommands/SHIP.md
Keep the starter kit generic.
Do not hard-code one business domain into the kit itself.
Each actual system should provide its own requirements through a project-specific spec under specs/.
The orchestration layer exists to ensure those requirements are routed through the correct lifecycle and skills, rather than handled as one-shot prompting.