The 7 Execution Phases

The 7 Execution Phases

Praetor processes codebases through a structured, linear pipeline of 7 distinct execution phases. This structured progression ensures that the initial code understanding is solid before specialized audits and Quality Council reviews begin.

---

📅 Phases Overview

Phase 0 ──► Phase 1 ──► Phase 2 ──► Phase 3 (MUST CONFIRM) ──► Phase 4 (Swarm) ──► Phase 5 (QC) ──► Phase 6 (Wrap-up)

---

🏛️ Deep Dive into the Pipeline

Phase 0: Source Resolution

• Goal: Prepare codebase entry points.
• Details: Unpacks the repository sources, resolves submodule structures, filters out irrelevant files (like compiled binaries, static icons, or package lockfiles), and prepares a clean, text-only source-code cache.

Phase 1: Technical Discovery

• Goal: Auto-detect tech stack, architecture, and configurations.
• Details: Scans config files, dependencies, and imports to detect:
  • Languages, databases, and core framework libraries.
  • Logging platforms (Winston, Sentry, Logback).
  • Observability platforms (Datadog, Prometheus, NewRelic).
  • Deployment and environment environments (Docker, Kubernetes).

Phase 2: Domain Mapping

• Goal: Construct logical boundaries.
• Details: Performs Domain-Driven Design (DDD) scans, grouping files into domain services, bounding contexts, and data flow layers. This serves as the blueprint that Judge 1 (Coverage) will use to audit agent output completeness.

Phase 3: Discovery Report & Gate (MUST CONFIRM)

• Goal: Validate and override parameters before swarming.
• Details: The system pauses and outputs a structured Discovery Report detailing all detected tools, scopes, and parameters. The user must provide a confirmation block (e.g., using CONDITIONAL_CONTINUE protocol) containing:
  • Answers to key discovery questions (e.g., Is WCAG compliance required?).
  • Run mode overrides (e.g., limiting audits to specific directories or high-priority items only).

Phase 4: Agent Swarm Analysis

• Goal: Dispersed parallel analysis.
• Details: Dispatches the specialized Tier 2, Tier 3, and Tier 4 agents (Security, Perf, SRE, Compliance) to execute their mandates. The agents write their findings in custom formats matching the target frameworks detected during Phase 1.

Phase 5: Quality Council Audit

• Goal: Eliminate hallucinations and guarantee citation accuracy.
• Details: The 4-Judge Quality Council audits all agent outputs. Any failed spec is sent back to the respective agent for regeneration with inline feedback.

Phase 6: Wrap-up & Release Gate Delivery

• Goal: Consolidate and package results.
• Details: Integrates all passed deliverables into structured folders organized by target audience. It generates the Unified Risk Register (remediation times, owners, priority levels) and the Main Release Gate checklist shell hook to prevent regressions.