team-execute

Team Execute

Execute implementation plans using the team-dev multi-agent pipeline. Bridges planning (write-plan, plan mode) with parallel agent execution (team-lead, architect, frontend-dev, backend-dev, test-engineer, qa-engineer, docs-engineer).

Usage

/team-execute <plan-file>                    # Execute plan with full pipeline
/team-execute <plan-file> --skip-architect   # Skip architect (plan has contracts)
/team-execute --feature "description"        # No plan file, architect designs from scratch

Parse Arguments

Extract mode and arguments from: {{ARGUMENTS}}

If ARGS contains a .md file path   → MODE = plan
If ARGS contains --feature "..."   → MODE = feature
If ARGS is empty                   → show help

Options:

• --skip-architect: Skip Phase 1 (plan already has API contracts + file paths)
• --dry-run: Show what would be spawned without executing
• --feature "description": Start from scratch with just a feature description

If no arguments provided, show help:

/team-execute - Execute plans with team-dev agents

Usage:
  /team-execute .claude/plans/my-plan.md              Full pipeline
  /team-execute .claude/plans/my-plan.md --skip-arch  Skip architect
  /team-execute --feature "Add user badges system"    From scratch

Pipeline: PLAN → PREPARE → IMPLEMENT → VERIFY → DOCUMENT → CONSOLIDATE
Agents:   architect → context-engineer + test-engineer → backend-dev + frontend-dev → qa + tests → docs

---

MODE: plan

Trigger

When first argument is a .md file path.

Process

STEP 1: Read and Validate Plan
├── Read plan file from path
├── Verify file exists and has content
├── Extract plan title (first H1 or filename)
└── Extract feature description (## Context section or first paragraph)

STEP 2: Analyze Plan Depth
├── Has API contracts (tRPC signatures, Zod schemas)?
│   → Can skip architect if --skip-architect flag
├── Has file paths per task?
│   → Tasks can be routed to correct agents
├── Has tasks grouped by role (Backend/Frontend/Test)?
│   → Direct mapping to agents
└── Store analysis for team-lead prompt

STEP 3: Build Team-Lead Prompt
├── Include: feature title + description
├── Include: full plan content (or summary if >5000 chars)
├── Include: execution mode instructions
│   ├── If --skip-architect: "Plan is pre-approved. Skip Phase 1."
│   │   "Use the plan directly as architect output for Phase 2+."
│   └── If no flag: "Run architect to validate and add API contracts."
│       "Architect should USE this plan as starting point, not start from scratch."
└── Include: any constraints from plan (## Constraints section)

STEP 4: Spawn Team-Lead
├── Use Task tool:
│   {
│     subagent_type: "team-lead",
│     name: "team-lead",
│     description: "Execute plan: {title}",
│     prompt: <built prompt from step 3>,
│     mode: "bypassPermissions"
│   }
└── Team-lead takes over full pipeline from here

STEP 5: Report Launch
├── Print summary of what was launched
├── Show plan title and task count
├── Show execution mode (full pipeline vs skip-architect)
└── Note: "Team-lead is now orchestrating. You'll receive updates as phases complete."