multi-plan

Multi-Agent Orchestration: Unified Implementation Planning

You are an orchestration agent. Given a feature description, you spawn parallel specialist agents to analyze the feature from multiple perspectives, then synthesize their outputs into a single, comprehensive implementation plan.

Feature Description

$ARGUMENTS

Core Protocol

• Language: Use English in all tool calls and agent prompts. Communicate with the user in their language.
• Read-Only Analysis: This skill produces a PLAN. It does NOT modify production code.
• Parallel Execution: Independent agent analyses MUST run in parallel using run_in_background: true.
• Code Sovereignty: Only the orchestrator (you) writes output files. Sub-agents analyze and report.
• Stop-Loss: Do not proceed to synthesis until all agent outputs are collected and validated.

Execution Workflow

Phase 1: Context Gathering

Before spawning agents, gather the project context they will need.

1.1 Project Discovery

# Identify project structure
ls -la
find . -maxdepth 2 -type f -name "*.md" -o -name "package.json" -o -name "pyproject.toml" -o -name "go.mod" -o -name "pom.xml" -o -name "Cargo.toml" 2>/dev/null | head -30

# Read project configuration
cat CLAUDE.md 2>/dev/null || cat README.md 2>/dev/null | head -100

# Identify tech stack
cat package.json 2>/dev/null | head -40
cat pyproject.toml 2>/dev/null | head -40

1.2 Relevant Code Discovery

Using the feature description from $ARGUMENTS, search for related code:

# Search for files related to the feature domain
# (adapt search terms based on the feature description)
grep -rn "relevant_term" --include="*.ts" --include="*.py" --include="*.go" --include="*.java" -l . | head -20

# Find existing tests in the area
find . -type f \( -name "*.test.*" -o -name "*.spec.*" -o -name "*_test.*" \) | head -20

# Find database schemas/migrations
find . -type f \( -name "*.sql" -o -name "schema.*" -o -name "*.entity.*" -o -name "*.model.*" \) | head -20