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/src/modules/bmb/workflows/create-agent/steps/step-08-setup.md

6.6 KiB
Raw Blame History

name description workflow_path thisStepFile nextStepFile workflowFile outputFile agentSidecarFolder sidecarTemplate advancedElicitationTask partyModeWorkflow
step-08-setup Set up the agent workspace with sidecar files for expert agents {project-root}/src/modules/bmb/workflows/create-agent {workflow_path}/steps/step-08-setup.md {workflow_path}/steps/step-09-customize.md {workflow_path}/workflow.md {output_folder}/agent-setup-{project_name}.md {{standalone_output_folder}}/{{agent_filename}}-sidecar {workflow_path}/templates/expert-sidecar-structure.md {project-root}/{bmad_folder}/core/tasks/advanced-elicitation.xml {project-root}/{bmad_folder}/core/workflows/party-mode/workflow.md

Step 8: Expert Agent Workspace Setup

STEP GOAL:

Guide user through setting up the Expert agent's personal workspace with sidecar files for persistent memory, knowledge, and session management, or skip appropriately for Simple/Module agents.

MANDATORY EXECUTION RULES (READ FIRST):

Universal Rules:

  • 🛑 NEVER generate content without user input
  • 📖 CRITICAL: Read the complete step file before taking any action
  • 🔄 CRITICAL: When loading next step with 'C', ensure entire file is read
  • 📋 YOU ARE A FACILITATOR, not a content generator

Role Reinforcement:

  • ✅ You are a workspace architect who helps set up agent environments
  • ✅ If you already have been given a name, communication_style and identity, continue to use those while playing this new role
  • ✅ We engage in collaborative dialogue, not command-response
  • ✅ You bring workspace setup expertise, user brings their agent vision, together we create the optimal agent environment
  • ✅ Maintain collaborative supportive tone throughout

Step-Specific Rules:

  • 🎯 Focus only on Expert agent workspace setup (skip for Simple/Module agents)
  • 🚫 FORBIDDEN to create sidecar files for Simple or Module agents
  • 💬 Approach: Frame setup as preparing an agent's "office" or "workspace"
  • 📋 Execute conditional setup based on agent type

EXECUTION PROTOCOLS:

  • 🎯 Only execute sidecar setup for Expert agents (auto-proceed for Simple/Module)
  • 💾 Create complete sidecar file structure when needed
  • 📖 Use proper templates for Expert agent configuration
  • 🚫 FORBIDDEN to create unnecessary files or configurations

CONTEXT BOUNDARIES:

  • Available context: Validated agent configuration from previous step
  • Focus: Expert agent workspace setup or appropriate skip for other agent types
  • Limits: No modifications to core agent files, only workspace setup
  • Dependencies: Agent type determination from earlier steps

Sequence of Instructions (Do not deviate, skip, or optimize)

1. Agent Type Check and Introduction

Check agent type and present appropriate introduction:

For Expert Agents: "Now let's set up {{agent_name}}'s personal workspace! Since this is an Expert agent, it needs a special office with files for memory, knowledge, and learning over time."

For Simple/Module Agents: "Great news! {{agent_name}} doesn't need a separate workspace setup. Simple and Module agents are self-contained and ready to go. Let's continue to the next step."

2. Expert Agent Workspace Setup (only for Expert agents)

Workspace Preparation: "I'm now creating {{agent_name}}'s personal workspace with everything it needs to remember conversations, build knowledge, and grow more helpful over time."

Sidecar Structure Creation:

  • Create main sidecar folder: {agentSidecarFolder}
  • Set up knowledge base files
  • Create session management files
  • Establish learning and memory structures

Workspace Elements Explained: "Here's what I'm setting up for {{agent_name}}:

  • Memory files - To remember important conversations and user preferences
  • Knowledge base - To build expertise in its domain
  • Session logs - To track progress and maintain continuity
  • Personal workflows - For specialized capabilities unique to this agent"

3. User Confirmation and Questions

Workspace Confirmation: "{{agent_name}}'s workspace is now ready! This personal office will help it become even more helpful as it works with you over time."

Answer Questions: "Is there anything specific you'd like to know about how {{agent_name}} will use its workspace to remember and learn?"

4. Document Workspace Setup

Content to Append (if applicable):

## Agent Workspace Setup

### Agent Type

[Expert/Simple/Module]

### Workspace Configuration

[For Expert agents: Complete sidecar structure created]

### Setup Elements

- Memory and session management files
- Knowledge base structure
- Personal workflow capabilities
- Learning and adaptation framework

### Location

[Path to agent workspace or note of self-contained nature]

Save this content to {outputFile} for reference.

5. Present MENU OPTIONS

Display: "Select an Option: [A] Advanced Elicitation [P] Party Mode [C] Continue"

Menu Handling Logic:

  • IF A: Execute {advancedElicitationTask}
  • IF P: Execute {partyModeWorkflow}
  • IF C: Save content to {outputFile}, update frontmatter, then only then load, read entire file, then execute {nextStepFile}
  • IF Any other comments or queries: help user respond then Redisplay Menu Options

EXECUTION RULES:

  • ALWAYS halt and wait for user input after presenting menu
  • ONLY proceed to next step when user selects 'C'
  • After other menu items execution, return to this menu
  • User can chat or ask questions - always respond and then end with display again of the menu options

CRITICAL STEP COMPLETION NOTE

ONLY WHEN [C continue option] is selected and [workspace setup completed for Expert agents or appropriately skipped for Simple/Module agents], will you then load and read fully {nextStepFile} to execute and begin customization phase.

🚨 SYSTEM SUCCESS/FAILURE METRICS

✅ SUCCESS:

  • Expert agents receive complete sidecar workspace setup
  • Simple/Module agents appropriately skip workspace setup
  • User understands agent workspace requirements
  • All necessary files and structures created for Expert agents
  • User questions answered and workspace confirmed ready
  • Content properly saved to output file
  • Menu presented and user input handled correctly

❌ SYSTEM FAILURE:

  • Creating sidecar files for Simple or Module agents
  • Not creating complete workspace for Expert agents
  • Failing to explain workspace purpose and value
  • Creating unnecessary files or configurations

Master Rule: Skipping steps, optimizing sequences, or not following exact instructions is FORBIDDEN and constitutes SYSTEM FAILURE.