Steps

description: Initialize agent team for an existing project

Initialize the agent team for an existing project so it can use sustained session-based development.

Project name: $ARGUMENTS

Steps

1. Validate project exists:

   • If no project name was provided, ask for one
   • Check if docs/projects/{project}/ exists
   • If not, ask if the user wants to run /init-project first

2. Create state directory and files:

   Create docs/projects/{project}/state/ with these files:

   state/session-log.md:

   # Session Log: {project}
   
   <!-- Context Keeper appends to this file at the end of each session -->
   

   state/decisions.md:

   # Decisions: {project}
   
   <!-- Key decisions made during development. Context Keeper maintains this file. -->
   
   | Date | Decision | Rationale | Agent |
   |------|----------|-----------|-------|
   

   state/requirements.md:

   # Requirements: {project}
   
   <!-- Current requirements. Business Analyst owns content, Context Keeper syncs. -->
   
   ## Functional Requirements
   
   _To be defined by Business Analyst._
   
   ## Non-Functional Requirements
   
   _To be defined by Business Analyst._
   
   ## Success Criteria
   
   _To be defined by Business Analyst._
   

   state/last-review.md:

   <!-- Last periodic review date. Context Keeper checks this at session start. -->
   Never reviewed
   

3. Add agent team reference to project CLAUDE.md (if one exists):

   • Check if the project has its own CLAUDE.md
   • If yes, append the agent team reference block (see below)
   • If no, note that the project will use the parent CLAUDE.md

   Agent team reference block:

   ## Agent Team
   
   This project uses the agent team defined in:
   `~/AI/ClaudeCodingProjectSetup/agents/`
   
   At session start, run `/start-session {project}` to activate the Context Keeper.
   
   Agent workflow: Context Keeper → BA → Planner → Implementer → Validator
   

4. Confirm to user:

   • List all files created
   • Explain that /start-session {project} is now available
   • Explain the agent workflow they'll use