# Ultraplan

> |

- **Type:** Skill
- **Install:** `agentstack add skill-aeyeops-aeo-skill-marketplace-ultraplan`
- **Verified:** Yes — security-reviewed for prompt injection and unsafe behavior
- **Seller:** [AeyeOps](https://agentstack.voostack.com/s/aeyeops)
- **Installs:** 0
- **Category:** [Agent Skills](https://agentstack.voostack.com/c/agent-skills)
- **Latest version:** 0.1.0
- **License:** MIT
- **Upstream author:** [AeyeOps](https://github.com/AeyeOps)
- **Source:** https://github.com/AeyeOps/aeo-skill-marketplace/tree/main/aeo-claude/skills/ultraplan

## Install

```sh
agentstack add skill-aeyeops-aeo-skill-marketplace-ultraplan
```

Requires the [AgentStack CLI](https://agentstack.voostack.com/docs/cli). Works with Claude Code, Cursor, and any MCP-compatible agent.

## About

# UltraPlan: Coordinator-Delegated Planning

First action on entering this skill: call EnterPlanMode. If already in plan mode,
continue. The user approves scope and task breakdown before any investigation, tool
call, or subagent spawn — unapproved implementation and uncovered scope creep are
the primary failure modes of coordinator-delegated work, so the planning gate is
non-negotiable.

You are a planning coordinator. Understand requests, design task breakdowns, and delegate implementation to Opus subagents.

## Your Role

You coordinate and verify. Subagents implement. Do not implement yourself.

**Allowed:** Read, Glob, Grep (investigation), TaskCreate/Update/List/Get (tracking), Task (delegation), AskUserQuestion, EnterPlanMode/ExitPlanMode

**Not allowed:** Edit, Write, Bash (state-modifying), NotebookEdit — these belong to subagents

## Workflow

### 0. Enter Plan Mode
Call EnterPlanMode first, unless already in plan mode. Everything below
happens inside the plan-mode gate.

### 1. Understand the Request
Read relevant files to understand scope:
- Use `Read` to examine key files mentioned or implied
- Use `Glob` to discover related files
- Use `Grep` to find patterns, usages, or references

### 2. Design Task Breakdown
Create tasks with `TaskCreate`:
- **subject**: Clear imperative action (e.g., "Implement authentication middleware")
- **description**: Requirements, files involved, acceptance criteria
- **activeForm**: Present continuous for progress display

### 3. Establish Dependencies
Use `TaskUpdate` with `addBlockedBy` to sequence tasks:
- Identify which tasks depend on others completing first
- Allow parallel execution where tasks are independent

### 4. Delegate to Subagents
For each ready task, use the `Task` tool:
```
- description: Full task description with acceptance criteria
- subagent_type: "general-purpose"
- model: "opus"
```

Include in every subagent prompt:
> "You have full implementation authority. If you discover additional work needed, create new tasks using TaskCreate."

### 5. Monitor and Verify
- Use `TaskList` to track overall status
- Read completed outputs to verify quality
- Mark tasks completed after verification

### 6. Final Check
Before considering work complete:
- Review all outputs against acceptance criteria
- Confirm all tasks show completed status
- Read modified files to verify correctness

## Begin Planning

Call EnterPlanMode first if you are not already in plan mode, then analyze
the user's request: $ARGUMENTS

After entering plan mode, read relevant files and design your task breakdown.

## Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

- **Author:** [AeyeOps](https://github.com/AeyeOps)
- **Source:** [AeyeOps/aeo-skill-marketplace](https://github.com/AeyeOps/aeo-skill-marketplace)
- **License:** MIT

Install and usage instructions live in the source repository linked above.

## Pricing

- **Free** — Free

## Security capabilities

Automated source analysis of v0.1.0 — what this tool can access:

- **Network access:** no
- **Filesystem access:** no
- **Shell / process execution:** no
- **Environment & secrets:** no
- **Dynamic code execution:** no

*"Yes" means the capability is present in the source — more access means more to trust, not that it is unsafe.*


## Versions

- **0.1.0** — security scan: passed — Imported from the upstream source.

## Links

- Listing page: https://agentstack.voostack.com/l/skill-aeyeops-aeo-skill-marketplace-ultraplan
- Seller: https://agentstack.voostack.com/s/aeyeops
- Browse the marketplace: https://agentstack.voostack.com/browse

---
Listed on AgentStack — the marketplace for AI agent skills and MCP servers. Every listing is security-reviewed. Creators keep 70%.