💡 摘要
PromptL 是用于在 Latitude 平台上创建和管理动态提示的语法指南。
🎯 适合人群
🤖 AI 吐槽: “看起来很能打,但别让配置把人劝退。”
风险:Low。建议检查:文件读写范围与路径穿越风险。以最小权限运行,并在生产环境启用前审计代码与依赖。
name: promptl description: PromptL syntax guide for writing prompts in the Latitude platform. This skill should be used when writing, reviewing, or editing PromptL prompts. Triggers on tasks involving creating prompts, configuring LLM parameters, using variables, conditionals, loops, chains, tools, agents, or any prompt engineering in Latitude. license: MIT metadata: author: latitude version: "1.0.0"
PromptL Syntax Guide
Comprehensive guide for writing PromptL prompts in the Latitude platform. PromptL is a versatile, human-readable language that simplifies defining and managing dynamic prompts for LLMs.
When to Apply
Reference these guidelines when:
- Writing new PromptL prompts for Latitude
- Configuring LLM models and parameters
- Using variables, conditionals, or loops in prompts
- Creating multi-step chains or agentic workflows
- Defining tools for function calling
- Referencing other prompts (snippets)
- Working with structured JSON output schemas
Syntax Categories
| Category | Purpose | Key Syntax |
|----------|---------|------------|
| Configuration | Define model, provider, and parameters | --- YAML block |
| Messages | Structure conversations | <system>, <user>, <assistant>, <tool> |
| Variables | Dynamic content | {{ variable }} |
| Conditionals | Dynamic flow control | {{ if }}, {{ else }}, {{ endif }} |
| Loops | Iterate over lists | {{ for item in list }}, {{ endfor }} |
| Chains | Multi-step prompts | <step> |
| Tools | Function calling | tools: in config |
| Agents | Autonomous workflows | type: agent |
| Snippets | Reusable prompts | <prompt path="..." /> |
| Content | Multi-modal content | <content-image>, <content-file> |
Quick Reference
Configuration Block
--- provider: OpenAI model: gpt-4o temperature: 0.7 top_p: 0.9 maxSteps: 20 ---
Message Tags
<system>- System instructions<user>- User messages<assistant>- Assistant responses<tool>- Tool interaction results<message role="...">- Generic message tag
Variables
{{ variable_name }}
{{ variable || "default" }}
{{ set myVar = "value" }}
Conditionals
{{ if condition }}
content
{{ else }}
alternative
{{ endif }}
Loops
{{ for item, index in items }}
{{ index }}: {{ item }}
{{ else }}
No items
{{ endfor }}
Chains (Multi-step)
<step as="result">
First step content
</step>
<step>
Use {{ result }} from previous step
</step>
Tools Configuration
tools: - tool_name: description: What the tool does parameters: type: object properties: param_name: type: string description: Parameter description required: - param_name
Agents
--- type: agent provider: OpenAI model: gpt-4o tools: - latitude/search agents: - agents/sub-agent-path maxSteps: 40 ---
Structured Output
schema: type: object properties: field_name: type: string description: Field description required: - field_name
Prompt References (Snippets)
<prompt path="relative/path/to/prompt" />
<prompt path="shared/policies" variable={{ value }} />
Content Types
<content-text>Plain text</content-text>
<content-image>{{ image_url }}</content-image>
<content-file mime="application/pdf">{{ file_data }}</content-file>
How to Use
Read the AGENTS.md file for the complete detailed guide with all syntax explanations, examples, and best practices.
Each section contains:
- Detailed explanation of the feature
- Correct usage examples
- Common mistakes to avoid
- Best practices and tips
优点
- 全面的语法覆盖
- 支持动态提示创建
- 促进多步骤工作流程
缺点
- 需要熟悉 YAML
- README 中示例有限
- 依赖外部文档
相关技能
免责声明:本内容来源于 GitHub 开源项目,仅供展示和评分分析使用。
版权归原作者所有 latitude-dev.
