ldy26098
/
BMAD-METHOD
mirror of https://github.com/bmad-code-org/BMAD-METHOD.git
1
0
Fork 0
Code Packages Projects Releases Wiki Activity
BMAD-METHOD/README-changes.md

9.3 KiB
Raw Blame History

Fork Changes: @jonahschulte/bmad-method

This document describes the customizations and enhancements in this fork compared to the upstream BMAD-METHOD repository.

Overview

This fork extends the official BMAD Method with production automation workflows focused on:

  • Quality assurance automation - Multi-stage validation pipelines
  • Token efficiency - 50-70% reduction through complexity-based routing
  • Autonomous processing - Unattended epic completion
  • Anti-vibe-coding enforcement - Mandatory gap analysis and code review
  • Smart routing - Micro stories skip unnecessary quality gates

Stats: ~26,000+ lines of additions across 122 files

Key Features Added

1. Super-Dev Pipeline (/super-dev-pipeline)

A comprehensive quality workflow that ensures stories are truly complete before marking done:

Pre-validation → Development → Post-validation → Code Review → Fixes → Done ✅

What it solves:

  • Prevents premature task completion ("stupid-dev" syndrome)
  • Catches partial implementations
  • Finds missed edge cases automatically
  • Runs adversarial code review before human review

Complexity-Based Routing (v1.3.0):

  • MICRO stories automatically skip steps 2 (pre-gap analysis) and 5 (code review)
  • STANDARD/COMPLEX stories use full 7-step pipeline
  • Early bailout checks prevent processing already-complete or invalid stories
  • Token savings: 50-70% for micro stories, 90% for early bailouts

Location: src/modules/bmm/workflows/4-implementation/super-dev-pipeline/

2. Story Pipeline v2.0 (/story-pipeline)

Single-session step-file architecture with 60-70% token savings:

Mode Token Usage Savings vs Legacy
Interactive ~25K 65%
Batch (YOLO) ~30K 58%

How it works:

  • Replaces 6 separate Claude CLI calls with 1 session
  • Just-in-time step loading
  • Checkpoint/resume capability
  • Role switching within session (SM → TEA → DEV → SM)

Automated steps:

  1. Initialize & load context
  2. Create story (SM role)
  3. Validate story (adversarial - must find 3-10 issues)
  4. ATDD test generation (TEA role)
  5. Implementation (DEV role)
  6. Code review (adversarial)
  7. Complete & commit
  8. Generate audit trail

Location: src/modules/bmm/workflows/4-implementation/story-pipeline/

3. Batch Super-Dev (/batch-super-dev) ⭐ Primary

Interactive batch workflow for processing multiple stories with full quality gates:

/batch-super-dev

→ Select stories: 1-3,5,7  (or "all")
→ Validates story files (creates missing, regenerates invalid)
→ Scores complexity for each story (micro/standard/complex)
→ Routes to appropriate pipeline (lightweight vs full)
→ Processes each with super-dev-pipeline
→ Reconciles story status after each
→ All stories complete! ✅

Features (v1.3.0):

  • Complexity-Based Routing ⭐ NEW - Smart pipeline selection
    • MICRO (≤3 tasks, low risk): Lightweight path, skips gap analysis + code review (50-70% token savings)
    • STANDARD (4-15 tasks): Full pipeline with all quality gates
    • COMPLEX (≥16 tasks or high-risk): Enhanced validation, suggest splitting
  • Smart Story Validation - Checks 12 required BMAD sections
  • Auto-Creation - Creates missing story files with codebase gap analysis
  • Auto-Regeneration - Regenerates invalid/incomplete stories
  • Sequential or Parallel execution modes
  • Story Reconciliation - Verifies checkboxes match implementation
  • Epic filtering - Process only specific epic's stories

Complexity Scoring Algorithm:

complexity_score = task_count + (high_risk_keywords × 5) + (medium_risk_keywords × 2)

MICRO:    task_count ≤ 3 AND score ≤ 5 AND file_count ≤ 5 AND no HIGH risk keywords
COMPLEX:  task_count ≥ 16 OR score ≥ 20 OR has HIGH risk keywords (auth, security, payment, etc.)
STANDARD: everything else

Risk Keywords:

  • HIGH (5 points): auth, security, payment, encryption, migration, database, schema
  • MEDIUM (2 points): api, integration, external, third-party, cache
  • LOW (0 points): ui, style, config, docs, test

Location: src/modules/bmm/workflows/4-implementation/batch-super-dev/

4. Gap Analysis (/gap-analysis)

Dev-time codebase validation that solves the batch planning staleness problem:

The Problem: Stories planned days ago become stale as the codebase evolves.

The Solution: Mandatory gap analysis before development:

  1. Scans current codebase
  2. Validates DRAFT tasks against reality
  3. Proposes refinements (add/modify/remove tasks)
  4. User approves with 6 options: Y/A/n/e/s/r

Location: src/modules/bmm/workflows/4-implementation/gap-analysis/

5. Push-All Workflow (/push-all)

Safe automated git operations with comprehensive safety checks:

  • Secret detection (API keys, credentials)
  • Large file warnings
  • Build artifact detection
  • Smart commit message generation
  • Auto-push with error handling

Location: src/modules/bmm/workflows/4-implementation/push-all/

7. Story Validation System

Multi-level story validation with LLM-powered verification:

Workflow Purpose
/validate-story Single story validation
/validate-story-deep Comprehensive deep validation
/validate-all-stories Batch validation for epic
/validate-all-stories-deep Deep validation for all stories
/validate-epic-status Epic status reconciliation
/validate-all-epics Full sprint validation

Location: src/modules/bmm/workflows/4-implementation/validate-*/

8. Sprint Status Management Scripts

Python/TypeScript utilities for sprint tracking:

Script Purpose
sprint-status-updater.py Update story statuses across epics
task-verification-engine.py LLM-powered task completion verification
llm-task-verifier.py Haiku-based task validation
validation-progress-tracker.py Track validation progress
add-status-fields.py Add status fields to story files

Features:

  • Bedrock/Claude API clients with rate limiting
  • File utilities for story parsing
  • Resumable validation runs

Location: scripts/lib/

9. BMAD Guide Skill

Automatic Claude Code skill for BMAD navigation:

# Auto-installed with BMAD
# Provides workflow selection guidance

What it does:

  • Phase detection (where am I?)
  • Project level routing
  • Workflow decision tree
  • Critical rules enforcement

Location: resources/skills/bmad-guide.md

10. Claude Code Commands

Pre-configured Claude CLI commands:

Command Description
batch-super-dev Run batch super-dev pipeline
story-pipeline Execute story pipeline
super-dev-pipeline Run super-dev with quality gates
validate-epic-status Validate epic status
validate-all-epics Validate all epics

Location: .claude-commands/

Package Changes

{
  "name": "@jonahschulte/bmad-method",
  "description": "...Enhanced with super-dev-pipeline and smart batching"
}

This fork is published to npm under the @jonahschulte scope for independent installation.

Modules Modified

BMM (BMAD Main Method)

  • Enhanced dev.agent.yaml with new workflow menu items
  • Enhanced sm.agent.yaml with new workflow menu items
  • Modified create-story workflow for requirements-focused planning
  • Enhanced dev-story workflow with gap analysis integration

BMGD (BMAD Game Dev)

  • Added parallel workflows for game development context
  • Enhanced game-dev.agent.yaml and game-scrum-master.agent.yaml

Installation

This fork can be installed via:

npm install @jonahschulte/bmad-method

Or used directly by cloning this repository.

Documentation Added

Document Purpose
docs/super-dev-mode.md Super-dev workflow guide
docs/autonomous-epic-processing.md Auto-epic documentation
docs/gap-analysis.md Gap analysis user guide
docs/gap-analysis-migration.md Migration from legacy
docs/workflows/SPRINT-STATUS-SYNC-GUIDE.md Sprint status management
docs/HOW-TO-VALIDATE-SPRINT-STATUS.md Validation how-to
FEATURE-SUMMARY.md Complete feature overview
INTEGRATION-NOTES.md Integration documentation
TESTING-GUIDE.md Testing procedures

Compatibility

  • Backwards compatible with upstream BMAD
  • No breaking changes to existing workflows
  • Users can opt-in to new features
  • All schema validation passing
  • All linting/formatting passing
  • All tests passing

Contributing Back

These features are candidates for upstream contribution:

  1. Gap Analysis - Solves a real problem with batch planning
  2. Story Pipeline v2.0 - Significant token savings
  3. Super-Dev Mode - Quality enforcement

See PR-DESCRIPTION.md and PR-STORY-PIPELINE.md for prepared PR descriptions.

Last Sync with Upstream

Date: 2026-01-07 Upstream Commit: 5c766577 (Add CNAME file) Merge Commit: 9df79392

Latest Fork Changes:

  • Date: 2026-01-07
  • Commit: 9bdf4894 (Critical complexity routing fixes)
  • Features: Complexity-based routing v1.3.0, multi-agent review integration

To sync with upstream:

git fetch upstream
git merge upstream/main