[mwbrooks]

Summary

This pull request gives developers who build this sample with a coding agent a short AGENTS.md with the guidance an agent can't work out from the code, and points every agent to the README's install steps for the Slack plugin.

• The guidance in each .claude/CLAUDE.md moves to AGENTS.md, which most coding agents read. Each CLAUDE.md is now one line, @AGENTS.md, so Claude Code reads the same file.
• Each AGENTS.md keeps only how we want agents to work: run lint and tests before finishing, our code standards, and at the root, keeping the implementations in parity. Commands, layout, and architecture are left for the agent to read from the project, so they can't go stale here.
• Each project folder has its own AGENTS.md, so the guidance still works for a developer who only has that folder.
• A Slack Plugin section in each AGENTS.md links to the README's "Claude Code, Codex, Cursor, and Other Coding Agents" section instead of repeating the install commands.
• Same shape as docs: add AGENTS.md for coding agents bolt-js-starter-agent#156.

Testing

• Each link's anchor matches the install section heading in the README next to it.
• The before-you-finish commands match the ones CI runs for this repo.
• The @AGENTS.md import is the same one checked with claude -p (Claude Code 2.1.304) on docs: add AGENTS.md for coding agents bolt-js-starter-agent#156.