legion

DEPRECATED compatibility shim; workflow-common-core is canonical for all conventions referenced here

Interprets --just-* and --skip-* command flags for /legion:build and /legion:review, validates flag combinations against rules, and resolves the matching team template from intent-teams.yaml. Use when the user passes intent flags to build or review commands, asks about flag combinations, or needs to filter agent teams by intent.

Dev-QA loop engine with structured feedback, fix routing, and user escalation for /legion:review

Executes wave-structured plans with personality-injected agents — parallel or sequential per CLI adapter

Maps all 48 Legion agents by division, capability, and task type for intelligent team assembly

Engine for /legion:map. Analyzes an existing codebase, generates CODEBASE.md for backward-compatible architecture context, and writes .planning/codebase/ index artifacts consumed by /legion:start, /plan, /build, /review, /status, and /quick.

Design system creation, UX research workflows, and three-lens design review cycles for design-focused phases

Bridge Codex requests to the Legion workflows in this repository when the user references Legion or legacy /legion:* commands.

Marketing campaign planning — campaign documents, content calendars, cross-channel coordination, and marketing-specific phase decomposition

Pre-mortem analysis and assumption hunting for stress-testing plans before execution

Research-first design discovery engine for /legion:explore

Pre-coding specification pipeline that produces structured spec documents through 5 stages

Lean core conventions and path contracts used by all /legion: commands

Milestone management — definition, completion with metrics, archiving, summary generation

Breaks down a roadmap phase into phased implementation plans grouped by dependency order, assigns specialist agents to each plan, and generates PLAN.md files. Core engine for /legion:plan. Use when the user wants to plan a project phase, create implementation plans, organize work into waves, or assign agents to tasks.

Adaptive project initialization questioning that captures vision, requirements, and preferences

Validates and enforces agent authority boundaries during wave execution and review panels

Cross-CLI dispatch infrastructure for routing work to external CLIs

Structured multi-pass code cleanup engine for removing noise, simplifying structure, improving naming, and normalizing conventions

Guided agent personality creation with schema validation and registry integration

Governance escalation tier with dynamic agent panels, deliberation, voting, and audit persistence

Tracks execution progress with STATE.md updates, ROADMAP.md progress, and atomic git commits per plan

GitHub integration — issue tracking, PR creation, milestone sync, and status readback via gh CLI

Claude Code hooks for automated Legion lifecycle events

Stores and retrieves task outcomes and agent performance across sessions via .planning/memory/OUTCOMES.md, scores recalled entries by recency and importance, and degrades gracefully when memory files are absent. Use when the user references past sessions, asks to recall previous results, records lessons learned, or is working within a Legion multi-agent workflow.

Multi-project portfolio management — registry, state aggregation, dependencies, agent allocation

Multi-pass specialized evaluators for deep code quality, UI/UX, integration, and business logic review

OWASP Top 10 and STRIDE threat modeling security review for code and architecture

Pre-ship quality gates, deployment verification, and canary monitoring for the /legion:ship command

Optional cross-domain conventions for design/marketing/specialized workflows

Optional GitHub interaction conventions shared across commands

Optional memory conventions shared across commands

Assembles a review panel of specialist agents matched to the code or artifact being reviewed, each evaluating against domain-specific rubrics, then synthesizes findings into a consolidated report. Called by review-loop for /legion:review. Use when the user needs multi-reviewer feedback, peer review simulation, rubric-based code evaluation, or expert critique from multiple perspectives.