ajianaz/codomate/bmad-party-mode-workflow

BMAD Party Mode Workflow

What It Does

Runs 6-persona BMAD analysis (Winston, Amelia, Mary, John, Paige, Tesla) for strategic decision-making on product/business concepts.

Workflow

Prerequisites

• bmad-party-mode-enhancement skill loaded (prerequisite)
• A coding agent available for subagent execution
• Research data already collected

Execution

Option A: Subagent via Coding Agent (preferred if fast model available)

# Run BMAD analysis via a coding agent subagent
${CODING_AGENT} run --model ${MODEL} --prompt "<full BMAD prompt with 6 personas>" --max-tokens 4000

CRITICAL:

• Use a fast model (${MODEL}) — slower/heavier models may time out on long 6-persona analysis
• --max-tokens 4000 minimum for 6 personas
• May still timeout if prompt is too long (>2000 words input)

Option B: Manual Synthesis (fallback)

If the coding agent times out or is unavailable:

1. Run each persona as a separate shorter subagent call (1-2 personas per call)
2. OR write synthesis directly based on research data using your own analysis

Agent Discussion Cross-Validation

After BMAD analysis, validate findings with other agents:

1. Business Analyst — send analysis to a business analyst agent for viability review
2. Devil's Advocate — delegate a critical review to a separate subagent for contrarian perspective
3. Collect feedback, cross-validate with BMAD findings

Output Format

Synthesis report with:

• 6-perspective BMAD table (consensus per question)
• Agent discussion cross-validation
• Consensus table with final verdict per dimension
• Risk matrix (probability × impact × mitigation)
• Clear recommendation with exit criteria

Pitfalls

• Model timeout on heavy analysis: BMAD Party Mode is resource-intensive. Use a fast model or split into smaller calls.
• Subagent delegation timeout: Long analytical tasks (>5 min) often timeout. Use for focused single questions only.
• Unregistered agents: Ensure cross-validation agents are properly configured before attempting delegation.
• Max concurrent subagents: Respect the platform's concurrent subagent limit (typically 2).

When to Use

• New product/concept feasibility analysis
• Market entry decisions
• Competitive landscape assessment
• Strategic pivot evaluation