devkit

SENTINEL DevKit

Agent-first development toolkit for the Sentinel ecosystem.

What's Inside

devkit/
├── extension/              VS Code extension (TypeScript, 29 source files)
│   ├── src/agents/         Agent orchestration, swarm spawner, model router
│   ├── src/kanban/         Task management board
│   ├── src/qa/             QA loop automation
│   ├── src/tdd/            TDD runner integration
│   └── src/panels/         Dashboard webview panel
│
├── hooks/                  Git hooks
│   ├── pre-commit.sh       Bash (Linux/macOS)
│   ├── pre-commit.ps1      PowerShell (Windows)
│   └── pre-exec-guard.ps1  Execution guard
│
├── prompts/                AI agent prompt templates
│   ├── reviewer.md         Code reviewer agent
│   ├── fixer.md            Automated fixer agent
│   ├── security-audit.md   Security auditor agent
│   └── financial-guard.md  Financial security guard
│
├── rules/                  Agent behavior rules
│   ├── rationalizations.md TDD rationalization patterns (anti-patterns to avoid)
│   └── clawdbot-security.md Agent security rules (OWASP Agentic Top 10)
│
├── skills/                 Reusable agent skills
│   ├── agent-security/     OWASP Agentic security audit
│   ├── two-stage-review/   Spec compliance + code quality review
│   ├── qa-fix-loop/        Reviewer → Fixer automated cycle
│   └── tdd-enforcement/    TDD Iron Law enforcement
│
├── specs/                  Specification templates
│   └── dot-templates.md    DOT flowchart templates for specs
│
├── ui/                     Standalone web dashboard
│   ├── index.html
│   ├── app.js
│   └── styles.css
│
└── memory/                 Memory system integration
    └── rlm-integration.md  RLM Memory Bridge docs

VS Code Extension

Full-featured development extension with agent orchestration, kanban board, QA automation, and TDD enforcement.

cd extension
npm install
npm run build

Press F5 in VS Code to launch. Pre-built VSIX available at extension/sentinel-devkit-0.1.0.vsix.

Extension Features

Feature	Source	Description
Agent Orchestrator	src/agents/	Multi-agent execution with model routing
Swarm Mode	src/agents/SwarmSpawner.ts	Parallel agent spawning
Security Scanner	src/agents/SecurityScannerAgent.ts	Automated security scanning
Kanban Board	src/kanban/	Task tracking with persistence
QA Loop	src/qa/	Automated reviewer → fixer cycle
TDD Runner	src/tdd/	Test-first enforcement
Dashboard	src/panels/	Unified webview panel

Development Methodology

The DevKit enforces a structured development process:

SPECIFICATION (what to build)
    Requirements → Design → Tasks → Human Review
         ↓
IMPLEMENTATION (how to build correctly)
    TDD Iron Law: RED → GREEN → REFACTOR
    (No production code without a failing test first)
         ↓
REVIEW
    Stage 1: Spec Compliance (does it match the spec?)
    Stage 2: Code Quality (is it maintainable?)
         ↓
MERGE

Agent Prompts

Pre-built prompt templates for AI coding agents:

• reviewer.md — Structured code review with severity levels and actionable feedback
• fixer.md — Automated issue resolution with minimal diff changes
• security-audit.md — OWASP-aligned security scanning for AI agent code
• financial-guard.md — Financial data protection and compliance checks

Skills

Reusable skill definitions that can be loaded by compatible AI agents:

Skill	Description
agent-security	OWASP Agentic Top 10 2026 audit checklist
two-stage-review	Split review into spec compliance + code quality
qa-fix-loop	Autonomous reviewer → issue → fixer → re-review cycle
tdd-enforcement	Enforces test-first development, blocks rationalization

Git Hooks

Install pre-commit hooks to enforce quality gates:

# Linux/macOS
cp devkit/hooks/pre-commit.sh .git/hooks/pre-commit
chmod +x .git/hooks/pre-commit

# Windows PowerShell
Copy-Item devkit\hooks\pre-commit.ps1 .git\hooks\pre-commit