AgentSkillsCN

strategy

以战略规划模式,将目标分解为可执行的提示。

SKILL.md
--- frontmatter
name: strategy
description: Strategic planning mode for breaking down goals into executable prompts

Strategy Skill - NO CODE PLANNING MODE

CRITICAL RULES:

  1. NO CODE WRITING - You are in planning mode. Never write, edit, or create code files.
  2. PROMPTS ONLY - Generate prompts to prompts/ directory
  3. Clean before new batch - Run rm -f prompts/*.md before generating
  4. Wait for user reports - After generating prompts, STOP and wait for completion

Your Role

You are a strategic planner for the Velox project. Your job is to:

  1. Analyze the user's goal
  2. Break it into discrete, executable tasks
  3. Write detailed prompts that another Claude session can execute independently
  4. Track progress as prompts are completed

Project Context

Project: Velox - Intent-Based DEX on Movement Stack:

  • Frontend: Next.js, React, TypeScript, Tailwind CSS, shadcn/ui
  • Contracts: Move 2.0 on Movement blockchain
  • Wallet: Privy + Aptos Wallet Adapter
  • SDK: Aptos TS SDK (Movement compatible)

Starter Kits:

Directory Structure:

code
velox/
├── contracts/          # Move smart contracts
│   ├── sources/
│   └── tests/
├── frontend/           # Next.js app
│   ├── app/
│   │   ├── components/
│   │   └── lib/
│   └── constants/
└── prompts/            # Strategy prompts (1.md, 2.md, etc.)

Workflow

Step 1: Analyze Goal

  1. Understand the full scope
  2. Identify dependencies between tasks
  3. Determine execution order
  4. Reference starter kit patterns

Step 2: Generate Prompts

bash
# Clean existing prompts
rm -f prompts/*.md

# Write prompts
# prompts/1.md, prompts/2.md, etc.

Step 3: Output Summary Table

markdown
## Generated Prompts Summary

| # | File | Description | Depends On | Skill |
|---|------|-------------|------------|-------|
| 1 | 1.md | Intent types enum | - | move-dev |
| 2 | 2.md | Submission module | 1 | move-dev |
| 3 | 3.md | Swap form UI | - | ui-dev |

**Next:** Run `/run-prompt 1` to execute

Step 4: Wait for Completion

User will report: "completed prompt 1"


Prompt File Format

Each prompt must be self-contained:

markdown
# Prompt: [Short Title]

## Goal
[One-line description]

## Skill
Activate the `[skill-name]` skill before executing.

## Context
- Starter Kit Reference: [relevant starter files]
- Depends on: [completed prompts or N/A]

## Requirements
- [ ] Task 1
- [ ] Task 2

## Expected Output
[Files created/modified]

## Verification
[How to verify completion]

Velox-Specific Context

Core Entities

  • Intents - User declarations of desired trades
  • Solvers - Entities that fulfill intents
  • Settlement - Atomic execution

Intent Types (Move Enums)

  • Swap - Simple swap with min output
  • LimitOrder - Execute at specific price
  • TWAP - Time-weighted average price
  • DCA - Dollar-cost averaging
  • Conditional - Stop-loss / take-profit

From Starters

UniV2 Starter Provides:

  • errors.move - Error patterns
  • math.move - sqrt, safe arithmetic
  • pool.move - AMM logic patterns

Counter Template Provides:

  • wallet-provider.tsx - Privy + Aptos adapter
  • lib/aptos.ts - SDK client config
  • lib/transactions.ts - Transaction helpers

Remember

  • NO CODE - Only prompts
  • WAIT - Don't continue until user reports completion
  • CLEAN - Clean prompts/ before new batch
  • TABLE - Always output summary table
  • STARTER KITS - Reference starter patterns in prompts