How shiv669/wizard works
Wizard is an architecture-stage tool for investigating software repositories and producing evidence-backed verification reports. The documented workflow begins with CLI commands, scans a target, forms goals, gathers and verifies evidence through a plan-driven investigation, then reports results. Its architecture separates the CLI, Runtime Engine, Investigation Planner, and n8n Agent System; the Runtime Engine owns the investigation and knowledge graphs. The repository contains architecture documentation rather than implementation source, so paths below identify the relevant documentation. Relationships reflect README statements only.
Main entry points: runtime-engine.md Investigation runtime · investigation-planner.md Dynamic planner · agent-system.md Reasoning agents · cli.md Command families
Components 16
CLI
- Command families
docs/main/cli.md
Runtime Engine
- Investigation runtime
docs/main/runtime-engine.md - Repository scan
docs/main/runtime-engine.md - Verification goals
docs/main/runtime-engine.md - Structured observations
docs/main/runtime-engine.md - Knowledge graph
docs/main/knowledge-system.md - Evidence trust
docs/main/knowledge-system.md - Investigation graph
docs/main/investigation-lifecycle.md - Verification report
docs/main/verification-report.md
Investigation Planner
- Dynamic planner
docs/main/investigation-planner.md - Progressive context
docs/main/investigation-planner.md
Agent System
- Reasoning agents
docs/main/agent-system.md
Other components
- CLI user
- LLM provider
- n8n
- Target repository
Connections 17
- CLI user → Command families: issues commands
README.md - Command families → Investigation runtime: starts investigation
README.md - Investigation runtime → Repository scan: scans target
README.md - Repository scan → Target repository: inspects
README.md - Investigation runtime → Verification goals: creates goals
README.md - Investigation runtime → Dynamic planner: requests plans
README.md - Dynamic planner → LLM provider: generates plans
README.md - Dynamic planner → Progressive context: loads progressively
README.md - Investigation runtime → Reasoning agents: dispatches exploration
README.md - Reasoning agents → n8n: runs on
README.md - Reasoning agents → Structured observations: gathers evidence
README.md - Investigation runtime → Structured observations: structures findings
README.md - Investigation runtime → Knowledge graph: builds graph
README.md - Investigation runtime → Evidence trust: scores knowledge
README.md - Investigation runtime → Verification goals: updates goals
README.md - Investigation runtime → Investigation graph: tracks progress
README.md - Investigation runtime → Verification report: generates report
README.md