Concepts Overview

The key concepts that make RStack SDLC work.

The core problem RStack solves

AI coding agents fail at software delivery for predictable reasons:

• They code before requirements are clear
• They over-edit unrelated files
• They pass work as “done” without testing
• They forget previous mistakes and repeat them
• They make risky actions (deploys, deletes) without asking

RStack addresses all of these by giving the agent a structured operating model: an explicit lifecycle, team roles, approval gates, and evidence requirements.

The team model

RStack organizes agents into three teams:

Orchestrator (team lead)
├── Builder team
│   ├── Backend specialists
│   ├── Frontend specialists
│   ├── DevOps specialists
│   ├── Data specialists
│   └── Docs specialists
└── Validator team
    ├── QA specialists
    ├── Security specialists
    ├── Architecture reviewers
    ├── Code reviewers
    └── Performance / accessibility reviewers

The Orchestrator routes work, manages the lifecycle, and is the only agent that communicates with the product owner (you).

Builders implement tasks within a scoped, gated context.

Validators review builder output with read-only tools. They cannot modify source files.

The lifecycle

sdlc_orchestrate → sdlc_start → sdlc_clarify → sdlc_plan → sdlc_spec
       ↓
sdlc_approve (requirements, architecture)
       ↓
sdlc_build_next → sdlc_validate → (repeat)
       ↓
sdlc_approve (release-readiness)
       ↓
sdlc_memory

Each phase produces durable artifacts stored under .rstack/runs/<run_id>/.

Asset types

Agents

Markdown files that define agent roles, tools, and operating contracts.

Skills

Reusable workflow instructions that can be included in task packets.

Plugins

Domain packs bundling agents, skills, and commands for specific areas.

Run State

Durable per-run artifacts: plan, specs, tasks, approvals, traceability, evidence.

Asset counts

The RStack package currently ships:

Asset type	Count
Agents	196
Skills	156
Prompts	36
Plugins	72

Pipeline routing

The orchestrator maps lifecycle tasks to the right SDLC agents:

Pipeline stage	Agents involved
Product clarification	00-environment, 01-transcript
Requirements	02-requirements, 04-planning, 05-jira
Architecture	06-architecture, 12-security-threat-model, 14-cost-estimation
Implementation	07-code
Testing	08-testing
Security review	12-security-threat-model, 13-compliance-checker
Documentation	03-documentation, 10-summary
Release readiness	09-deployment, 10-summary, 11-feedback-loop

Each generated task includes routing metadata:

{
  "pipeline_agents": ["agent.02-requirements", "agent.04-planning"],
  "specialists": ["agent.02-requirements", "plugin.backend-development"]
}