Steven/system-prompts-and-models-of-ai-tools
1
0
Fork 0
mirror of https://github.com/x1xhlol/system-prompts-and-models-of-ai-tools.git synced 2026-02-07 15:30:52 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

添加总结

Browse Source 
添加总结
This commit is contained in:
tycon
2025-10-14 22:04:51 +08:00
parent c87083d594
commit 60ddd120c4
1067 changed files with 134118 additions and 10742 deletions
Download Patch File Download Diff File Expand all files Collapse all files

12
docs/en/amp/index.md
View File

@@ -19,4 +19,14 @@ The system prompt used by Amp is tuned to Sonnet 4.x and has other LLMs register
"amp.url": "https://ampcode.com/",
"amp.gpt5": true
}
```
```
## Summary of Product Tool Documents
This directory contains system prompts designed for the AI coding agent "Amp". Built by Sourcegraph, Amp is designed to help users with software engineering tasks. The files in this directory demonstrate how Amp is configured and optimized for different underlying large language models.
- **`claude-4-sonnet.md`**: This is the system prompt configured for Amp, targeting Anthropic's Claude Sonnet 4 model. It defines in detail Amp's agent behavior, task management (`todo_write` tool), code conventions, and communication style. A core feature is the frequent use of the `oracle` tool, an expert consultant played by another LLM (in this case, GPT-5), used for guidance in planning, reviewing, and debugging complex tasks.
- **`gpt-5.md`**: This is the system prompt configured for Amp, targeting OpenAI's GPT-5 model. This version also defines Amp's agent behavior but places more emphasis on a Parallel Execution Policy, rapid context understanding, and strict Guardrails. It also mentions the use of an `oracle` (possibly played by Claude Sonnet 4 here) and other sub-agents (`Task`, `Codebase Search`) to complete tasks collaboratively.
In summary, the `amp` directory showcases a flexible, multi-model collaborative AI agent architecture by providing customized system prompts for different LLMs. It utilizes a primary model (like Claude Sonnet 4) to execute tasks while using another powerful model (like GPT-5) as an "oracle" tool to provide expert advice, thereby achieving more powerful and reliable programming assistance capabilities.