Multi-AI Planning

Overview

multi-ai-planning creates comprehensive, agent-executable plans for complex objectives using verification-first approach, hierarchical task decomposition, and rigorous quality validation.

Purpose: Transform objectives into structured, validated plans that agents can execute reliably

Pattern: Workflow-based (6-step sequential process)

Key Innovation: Plans are schema-validated, quality-scored (≥90/100), and designed for multi-agent parallel execution with built-in verification and checkpoint systems.

Core Principles (validated by Claude + Gemini + Codex):

1. Verification-First: Define success criteria before implementation
2. Hierarchical Decomposition: Break down to atomic, executable steps
3. Explicit Dependencies: Map all dependencies, identify parallel opportunities
4. Quality Gates: Multi-layer validation before execution
5. Safe Checkpoints: Git tags + patches for resumption and rollback

---

When to Use

Use multi-ai-planning when:

• Planning complex features (>8 tasks, multiple components)
• Multi-step migrations or refactorings
• Security implementations requiring verification
• System integrations with many touch points
• Projects requiring multi-agent coordination
• When parallel execution optimization is important
• Quality and verification are critical

When NOT to Use:

• Simple tasks (<3 steps)
• Ad-hoc exploratory work
• Quick prototyping without formal structure

---

Prerequisites

Required

• Clear objective (know what you want to accomplish)
• Time for planning (1-3 hours depending on complexity)

Optional (for enhanced planning)

• multi-ai-research skill (for research phase)
• Gemini CLI (npm install -g @google/gemini-cli) - web research
• Codex CLI (npm install -g @openai/codex) - code pattern research

Understanding

• Familiarity with the objective domain
• Basic understanding of agent coordination (see references/task-tool-coordination.md)

---

Planning Workflow

Step 1: Objective Analysis & Research

Analyze the objective and optionally research best practices, patterns, and approaches before creating the plan.

Purpose: Ground plan in proven patterns and understand problem space

When to Use: Always (research optional but recommended for unfamiliar domains)

Inputs:

• Clear objective statement
• Context about current system (if applicable)
• Success criteria (what "done" looks like)

Process:

1. Clarify Objective:

   Objective: [Specific statement of what needs to be accomplished]
   
   Context:
   - Current state: [What exists now]
   - Desired state: [What should exist]
   - Constraints: [Time, resources, compatibility]
   
   Success Criteria:
   1. [Specific, measurable criterion]
   2. [Criterion 2]
   3. [Criterion 3]
   

2. Determine Plan Type:

   • Feature development (new functionality)
   • Refactoring (improve existing code)
   • Migration (move to new technology/pattern)
   • Security implementation (add security features)
   • Integration (connect systems)
   • Research project (investigate and document)

3. Optional: Research Domain (recommended for complex/unfamiliar):

   If using multi-ai-research skill:

   Use multi-ai-research for "Research [domain] best practices and implementation patterns"
   

   This will provide:

   • Official documentation insights (Claude)
   • Latest web best practices (Gemini)
   • GitHub implementation patterns (Codex)
   • Multi-source validation

   Manual research (if multi-ai-research not available):

   • Search web for "[domain] best practices 2024-2025"
   • Review official documentation
   • Find 3-5 example implementations
   • Document findings in research-notes.md

4. Analyze Approach Options: Based on research (if done), identify 2-3 possible approaches:

   ## Approach Options
   
   ### Option A: [Approach name]
   - **Description**: [How it works]
   - **Pros**: [Benefits]
   - **Cons**: [Drawbacks]
   - **Complexity**: Low/Medium/High
   - **Risk**: Low/Medium/High
   
   ### Option B: [Alternative]
   [Same format]
   
   ### Recommended: [Choice] because [rationale]
   

5. Define Initial Scope:

   ## Scope Definition
   
   **Included**:
   - [Component/area in scope]
   - [Component 2]
   
   **Excluded**:
   - [What we're NOT doing]
   - [Out of scope item 2]
   
   **Rationale**: [Why this scope is appropriate]
   

Outputs:

• Objective clarification document
• Plan type identified
• Research findings (if research conducted)
• Approach recommendation with rationale
• Initial scope definition

Validation:

• Objective is specific and clear
• Success criteria are measurable
• Plan type identified
• Approach chosen with rationale
• Scope defined (in/out)
• Research conducted (if needed)

Time Estimate: 30-60 minutes (research adds 30-60 min if used)

Next: Proceed to Step 2

...