Skip to content

agent-prompting-guide.md

Latest commit

 

History

History
66 lines (43 loc) · 3.03 KB

agent-prompting-guide.md

File metadata and controls

66 lines (43 loc) · 3.03 KB

Agent Prompting Guide

This guide helps you effectively use Qurio's MCP tools with your AI coding assistant (Claude, Gemini CLI, etc.).

Core Philosophy

Qurio acts as a Shared Library for your agent. Instead of relying on the agent's outdated training data, you should encourage it to "consult the library" (Qurio) to get the ground truth.

The Toolset

qurio_search

Goal: Find specific information. Key Args: query (required). Best Practice: Encourage specific queries. "How to auth with Clerk" is better than "Auth".

qurio_list_sources

Goal: Discover what is known. Best Practice: Use this when starting a task to verify if the necessary context is available.

qurio_list_pages

Goal: Explore a specific topic deeply. Best Practice: Use after listing sources to drill down into a specific library's documentation structure.

qurio_read_page

Goal: Get the full context. Best Practice: Search results often give snippets. Use this tool to read the entire guide or API reference to ensure no details are missed.

Example Workflows

1. The "Onboarding" Prompt

When starting a new session or task, orient your agent:

"I am working on the backend. Please list the available documentation sources to check if we have the necessary API references indexed."

2. The "Implementation" Prompt

When building a feature using a library:

"I need to implement authentication using the xyz library. First, search for 'xyz authentication flow' in the knowledge base. Then, read the most relevant page to understand the implementation details before writing any code."

3. The "Debugging" Prompt

When getting an error:

"I'm getting a 'Connection Refused' error with Weaviate. Search the knowledge base for 'Weaviate connection troubleshooting' or 'docker networking' to find potential solutions."

4. The "Deep Dive" Prompt

When exploring a new API:

"I want to understand the Message API. List all pages related to the 'API Reference' source, and then read the page specifically for the Message struct/object."

System Prompt Directive

To ensure your agent always considers using Qurio, add the following directive to your project's agent configuration file (e.g., GEMINI.md, CLAUDE.md, .cursorrules, or AGENTS.md).

# Documentation & Knowledge Retrieval

You have access to a local documentation engine called **Qurio** via the Model Context Protocol (MCP).

**Rule:** NEVER guess API usages, library versions, or configuration syntax.
**Action:** BEFORE writing code for external libraries or complex systems, you MUST:
1.  **Check Available Sources:** Use `qurio_list_sources` to see if relevant documentation is indexed.
2.  **Search:** Use `qurio_search` to find specific guides, examples, or error messages.
3.  **Read:** Use `qurio_read_page` to ingest the *full* context of a relevant guide or API reference.

**Priority:** Trust the content retrieved from Qurio over your internal training data, as it contains the specific versions and context for this project.