Agent Plan Writing
Parent: AI Agent Ecosystems · researched 2026-05-25· 12 sources · 10 concepts · skill agent-plan-writing
Agent plan writing is the discipline of designing execution plans for AI agent workflows. The harness matters more than the model. Agent completion rates depend more on action-space design, context en
Overview
- Agent plan writing is the discipline of designing execution plans for AI agent workflows. The harness matters more than the model. Agent completion rates depend more on action-space design, context engineering, and orchestration patterns. [source]
Output format
- When this skill activates, produce a markdown agent plan containing: [source]
- Workflow Overview - what the system does, which orchestration pattern, and why [source]
- Agent Roster - each agent's role, model, tools, and context scope [source]
- Orchestration Graph - how agents coordinate [source]
- Context Budget - token allocation per agent [source]
- Safety Constraints - permission boundaries, output validation, human-in-the-loop gates [source]
- Evaluation Plan - what to trace, quality metrics [source]
- Failure Handling - per-pattern failure modes and recovery strategies [source]
Orchestration patterns
- Five patterns dominate production agent systems: [source]
- Fan-Out: Parallel execution of independent subtasks. Coordinator dispatches to N agents simultaneously. [source]
- Pipeline: Sequential chain where each stage requires the prior stage's output. [source]
- Supervisor: A supervisor agent decomposes the task, delegates to specialists, and synthesizes results. The 2026 production default. [source]
- Debate: Multiple agents reason independently, then argue toward convergence. [source]
- Swarm: Dynamic spawning of agents based on workload. [source]
Context window budget planning
- Agents consume ~7x more tokens than standard chat sessions. Plan token budgets explicitly. [source]
- Budget allocation: [source]
- System prompt: 500–2,000 tokens. Cached input costs 10–25% of normal. [source]
- Tool schemas: Each MCP tool adds 100–500 tokens to context. [source]
- Working memory: Reserve 30–50% of context for conversation/reasoning accumulation. [source]
- Output headroom: Reserve 15–25% for the agent's response generation. [source]
Children
- Multi-Agent Orchestration Planning (frontier)
- Subagent Prompt Crafting (frontier)
- Context Window Budgeting (frontier)
- Safety Guardrails Design (frontier)
- Agent Evaluation Planning (frontier)
Frontier under this node: Agent Evaluation Planning, Context Window Budgeting, Multi-Agent Orchestration Planning, Safety Guardrails Design, Subagent Prompt Crafting