BMAD-METHOD/docs/super-dev-mode.md

10 KiB
Raw Blame History

title description
Super-Dev Mode - Comprehensive Quality Workflow Multi-stage validation workflow ensuring stories are truly complete through pre-validation, development, post-validation, and code review

Super-Dev Mode: Comprehensive Quality Workflow

TL;DR

Super-Dev Mode ensures stories are truly complete through multi-stage validation:

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

No more "stupid-dev" where tasks get checked off without actually being complete!

The Problem with "Stupid-Dev"

Standard dev-story Flow

1. Read tasks
2. Implement features
3. Check off tasks
4. Mark story "review"
5. ??? (Hope it's actually done)

What goes wrong:

  • Dev agent marks tasks complete prematurely
  • Partial implementations claimed as "done"
  • Missing edge cases not caught
  • Code issues discovered later in manual review
  • Story bounces back for rework

Super-Dev Flow

1. PRE-DEV GAP ANALYSIS ✅
   ↓ Validate tasks match codebase

2. DEVELOPMENT ✅
   ↓ Implement features

3. POST-DEV GAP ANALYSIS 🆕
   ↓ Verify nothing was missed

4. ADVERSARIAL CODE REVIEW 🆕
   ↓ Catch issues, security, quality problems

5. FIX REVIEW FINDINGS 🆕
   ↓ Address all problems

6. MARK DONE ✅
   ↓ Only when truly complete

What this prevents:

  • No premature task completion
  • Catches partial implementations
  • Finds missed edge cases
  • Code review happens automatically
  • Story only "done" when verified complete

How It Works

Stage 1: Pre-Dev Gap Analysis

Same as standard dev-story Step 1.5:

  • Scans codebase
  • Validates tasks against reality
  • Proposes task updates
  • User approves

Stage 2: Development

Standard implementation cycle:

  • Follows tasks sequentially
  • Red-green-refactor (TDD)
  • Runs tests
  • Updates story file

Stage 3: NEW - Post-Dev Gap Analysis

This is the game-changer!

After all tasks are checked off:

<step n="9.5" goal="Post-development gap analysis" if="super_dev_mode enabled">
  <critical>🔍 VERIFY COMPLETION - Did we actually build everything?</critical>

  <action>Re-scan codebase with original requirements</action>
  <action>Compare what should exist vs what actually exists</action>

  <check if="gaps found">
    <output>⚠️ POST-DEV GAPS DETECTED!

      Tasks marked complete but code missing or incomplete:
      {{list_of_gaps}}

      Adding missing work to task list...
    </output>

    <action>Add new tasks for missing work</action>
    <action>GOTO step 5 - Continue implementation</action>
  </check>

  <check if="no gaps">
    <output>✅ Post-dev validation passed - all work verified complete</output>
    <action>Continue to code review</action>
  </check>
</step>

Stage 4: NEW - Adversarial Code Review

Auto-run code review workflow:

<step n="9.6" goal="Adversarial code review" if="super_dev_mode enabled">
  <critical>👀 INDEPENDENT REVIEW - Catch issues before human review</critical>

  <action>Run code-review workflow automatically</action>
  <action>Generate review findings</action>

  <check if="critical issues found">
    <output>🚨 CODE REVIEW FOUND CRITICAL ISSUES:

      {{list_of_critical_issues}}

      Auto-fixing issues...
    </output>

    <action>Add review findings as tasks</action>
    <action>GOTO step 5 - Implement fixes</action>
  </check>

  <check if="only minor issues found">
    <output> CODE REVIEW FOUND MINOR ISSUES:

      {{list_of_minor_issues}}
    </output>

    <ask>Auto-fix minor issues? [Y/n]:</ask>

    <check if="user approves">
      <action>Add review findings as tasks</action>
      <action>GOTO step 5 - Implement fixes</action>
    </check>
  </check>

  <check if="no issues found">
    <output>✅ Code review passed - no issues found</output>
    <action>Continue to completion</action>
  </check>
</step>

Stage 5: Mark Done

Only after passing ALL validation:

✅ Pre-dev gap analysis passed
✅ Development completed
✅ Post-dev gap analysis passed
✅ Code review passed
✅ All issues fixed

→ NOW mark story "review"

Configuration

Enable Super-Dev Mode

Add to _bmad/bmm/config.yaml:

# Super-Dev Mode (comprehensive quality workflow)
super_dev_mode: true  # Default: false

# Super-Dev Options
super_dev:
  post_dev_gap_analysis: true    # Re-validate after implementation
  auto_code_review: true          # Run code review automatically
  auto_fix_minor_issues: false    # Auto-fix minor review findings without asking
  fail_on_gaps: false             # Fail hard if post-dev gaps found (vs auto-fix)

Per-Story Override

In story file frontmatter:

---
super_dev_mode: true
---

Usage

Enable for All Stories

# Edit config
vim _bmad/bmm/config.yaml

# Add:
super_dev_mode: true

# Now all dev-story runs use super-dev mode

Enable for One Story

# Edit story file
vim docs/sprint-artifacts/1-2-auth.md

# Add to frontmatter:
super_dev_mode: true

# Run dev-story as normal
/dev-story

Batch Enable

# Enable for all stories in epic 2
for story in docs/sprint-artifacts/2-*.md; do
  # Add super_dev_mode to frontmatter
  sed -i '1a super_dev_mode: true' "$story"
done

Examples

Example 1: Post-Dev Gap Catches Missing Test

Story: "Implement user authentication"

Tasks after development:
[x] Create auth service
[x] Implement login endpoint
[x] Write tests

POST-DEV GAP ANALYSIS:
✅ auth.service.ts exists
✅ login endpoint exists
❌ Tests claim 90% coverage but only 60% actual
❌ Edge case tests missing

ACTION: Add missing tests to task list, continue development

RESULT: Story truly complete with full test coverage

Example 2: Code Review Catches Security Issue

Story: "Add password reset feature"

Development complete, all tasks checked.

CODE REVIEW:
❌ CRITICAL: Password reset tokens not expiring
❌ MEDIUM: Error messages leak user existence
✅ All other checks passed

ACTION: Add security fixes to task list, continue development

RESULT: Secure implementation, no vulnerabilities

Example 3: Everything Passes First Try

Story: "Update dashboard UI"

Tasks completed perfectly.

POST-DEV GAP ANALYSIS: ✅ All work verified
CODE REVIEW: ✅ No issues found

RESULT: Story marked "review" immediately

Performance Impact

Stage Time Value
Pre-dev gap analysis 5-10s Prevents duplicate work
Development Normal Standard implementation
Post-dev gap analysis 5-10s Catches incomplete work
Code review 30-60s Finds quality/security issues
Total overhead 40-80s Prevents hours of rework

ROI: Spending 1 minute extra per story prevents hours of:

  • Rework from incomplete implementations
  • Bug fixes from missed edge cases
  • Security patches from vulnerabilities
  • Review cycles from quality issues

Comparison

Standard Dev-Story

Time: 30 minutes dev work
Quality: ??? (unknown until manual review)
Rework: High probability
Cycle time: Long (multiple review iterations)

Super-Dev Mode

Time: 31 minutes (30 dev + 1 validation)
Quality: High (multi-stage verified)
Rework: Minimal (issues caught early)
Cycle time: Short (passes review first time)

Best Practices

1. Enable for Complex Stories

Stories with:

  • Security implications
  • Complex business logic
  • Multiple integrations
  • High test coverage requirements

2. Disable for Trivial Stories

Stories like:

  • Documentation updates
  • Simple UI tweaks
  • Configuration changes

3. Use Auto-Fix for Minor Issues

super_dev:
  auto_fix_minor_issues: true

Saves time on:

  • Linting issues
  • Formatting problems
  • Simple refactoring

4. Batch Process with Caution

Super-dev mode increases time per story. For batch processing:

  • Enable for critical stories only
  • Or run overnight batch processing
  • Or accept longer processing time for quality

Troubleshooting

"Post-dev gap analysis found missing work every time"

Cause: Dev agent marks tasks complete prematurely

Solution: This is EXACTLY what super-dev catches! Let it add the missing work and continue.

"Code review fails every story"

Cause: Common coding issues not caught during development

Solution: This is working as designed. Review the patterns and add to "Dev Agent Guardrails" in future stories.

"Super-dev mode takes too long"

Cause: Multiple validation stages add time

Solution:

  • Disable for trivial stories
  • Use auto_fix_minor_issues: true
  • Accept quality over speed trade-off

"Post-dev analysis conflicts with pre-dev analysis"

Cause: Codebase changed during development (other stories completed)

Solution: This is normal for batch planning. Post-dev analysis adapts to evolved codebase.

FAQ

Q: Does this replace manual code review?

A: No. This catches obvious issues automatically. Human review still needed for:

  • Architecture decisions
  • Business logic correctness
  • UX/design decisions

Q: Can I use post-dev gap analysis without code review?

A: Yes:

super_dev:
  post_dev_gap_analysis: true
  auto_code_review: false

Q: Does this work with custom workflows?

A: Yes, if your custom dev-story preserves the hook points for gap analysis and code review.

Q: What if I disagree with code review findings?

A: You can reject auto-fixes and continue. The findings are logged for human review.

Implementation Status

  • Pre-dev gap analysis - Available in BMAD v6.0.0-alpha.22+
  • 🚧 Post-dev gap analysis - Planned for v6.0.0-alpha.23
  • 🚧 Auto code review - Planned for v6.0.0-alpha.23
  • 🚧 Super-dev mode config - Planned for v6.0.0-alpha.23

Contributing

Want to help implement super-dev mode?

  1. Post-dev gap analysis implementation
  2. Code review integration
  3. Config system for super-dev options
  4. Documentation and examples

See CONTRIBUTING.md for guidelines.


Super-Dev Mode: Because "done" should mean DONE, not "hope so"