admin: Add AI coding assistant configuration

[lgritz]

Adds agent configuration supporting Claude Code, Cursor, GitHub Copilot, OpenAI Codex, and Opencode. A single AGENTS.md and contents of .agents subdirectory serve as the canonical source for instructions, skills, and other things that can be used across multiple tools. To avoid duplication, tool-specific files are, whenever possible, either symbolic links or thin wrappers referring to canonical areas. Includes per-directory .gitignore files to prevent accidental commits of per-user settings and session data.

This initial commit includes two skills that I wrote and have used for the last couple monthly releases: release-notes-update and prepare-patch-release. I assume that we will add more over time as we discover tasks that we want to make standard skill automation for.

Note that our use of symlinks is a little troublesome on Windows, and assumes that Windows-based developers will (a) enable Windows Developer Mode, and (b) git config --global core.symlinks true, both prior to cloning the repo. I think that if you don't do that, Windows developers will get separate copies of those files rather than links, in their local tree. But even that doesn't really matter for people not changng these very files.

Notes:

• I don't use Windows, so if any of my assumptions here are wrong or put an undue burden on Windows users, please speak up and offer alternative suggestions or ask me to find a different approach.)
• I only use Claude Code and a bit of Opencode myself, so if I've done anything very wrong for Copilot, Cursor, or Codex, please make corrections. In an ideal world, over time, more of these will coalesce on some common standards so that there will be even less repetition (like, many of them already use AGENTS.md as the one name for general instructions, and they often have compatible skill files, though they all seem to want them stored in different places currently, hence the links).

Assisted-by: Claude Code / claude-sonnet-4-6

The AGENTS.md, Claude skills, and general scheme are entirely from me. I used Claude to help me establish the setup for the other tools I don't happen to use.

[lgritz]

Anybody have opinions about this? I'm especially interested in hearing from anybody who juggles multiple coding agents to give a sanity check for whether I'm on the right track with how to set up the directories and links to accommodate several tools and have them share as much as possible with minimal repetition.

[lgritz]

Take 2: minimal checked-in components, script to do per-tool setup.
(Thanks to @meshula for the suggestion.)

Change of strategy. Instead of checking in links and other niceties
specific to tools, ONLY check into the repo:

• AGENTS.md - general repo-level agent instructions
• .agents/ - directory with skills, etc., shared by all tools

A script can be invoked by users to do the local setup of links or
references to allow them to use a particular tool. For example,

.agents/setup-agent claude

Currently supported tools by this script include Claude Code,
Codex, Cursor, GitHub Copilot, and OpenCode.