projects: add server instructions

[tmelliottjr]

This PR updates the tool/argument descriptions for projects related tools to better enable tool selection and project item filter.

Additionally, this PR introduces pagination for list_* operations to enable effective querying.

[Copilot]

Pull Request Overview

This PR enhances the GitHub Projects V2 integration by adding advanced pagination support, improving response structures, refining query capabilities, and updating test coverage to match the new implementation. The changes focus on providing more robust pagination handling and clearer field filtering capabilities.

Key changes:

• Added cursor-based pagination support (after, before parameters) with pageInfo in responses
• Changed response format to include pageInfo metadata alongside data arrays
• Updated field filtering to exclude special/system data types from responses
• Refactored field ID handling from array to comma-separated string format

Reviewed Changes

Copilot reviewed 6 out of 6 changed files in this pull request and generated 2 comments.

File	Description
pkg/github/projects.go	Core implementation: adds pagination options extraction, pageInfo builder, filterSpecialTypes function, updated response structures with pageInfo, and comprehensive query documentation
pkg/github/projects_test.go	Test updates: modified mock data to include required fields (node_id), updated assertions to validate new response structure with pageInfo, simplified field query parameter handling
pkg/github/toolsnaps/*.snap	Snapshot updates: documents new pagination parameters and updated descriptions for tool schemas
README.md	Documentation updates: adds pagination parameters and extensive query syntax documentation for projects tools

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[kerobbi]

The technical improvements here lgtm! The only issues I can see are related to descriptions/prompts being quite verbose, and overly focused on technical details.

I have added a couple of comments just for better clarity, although they are all related to the same root concern.

[kerobbi]

Left some comments mainly regarding reducing/optimising context usage and avoiding duplicate instructions.

I also suggested removing a lot of the query related instructions, as in general we've been finding that the model is able to construct pretty good queries using the GitHub syntax. I am curious, have you found that the model has not been able to construct queries well? If so, was it happening even for simple queries or for more complex projects ones?

[stephenotalora]

Thanks, @kerobbi! I appreciate the feedback, @tmelliottjr is out until Monday, but I’m reviewing your comments in the meantime and will address them accordingly 🙇🏼

[stephenotalora]

@tmelliottjr just an fyi, recent push is to address a number of conflicts with main

[JoannaaKL]

Why is that different than the default?

[tmelliottjr]

I observed the models often using per_page: 100. Depending on the content in the payload and how many/what types of fields are requested this can result in some incredibly large response payloads. This was in service of trying to minimize that.

This along with encouraging the correct usage of query will hopefully reduce scenarios where truncation occurs.

[JoannaaKL]

Why do we even need to dedupe by id? If we're listing items by id they should contain unique items already.

[tmelliottjr]

Closed in favor of #1393