feat: add automated AGENTS.md generation for AI agents

[pkale]

This change introduces automated generation of AGENTS.md, a comprehensive documentation resource designed specifically for AI agents working with Aurora DSQL. The implementation ensures the documentation stays current through automated regeneration whenever source files change.

Key additions include:

• GitHub Actions workflow for automatic AGENTS.md regeneration
• agents-md.config.ts configuration covering all documentation files
• agents.md with custom AI agent-focused content and setup guides
• AGENTS_SETUP.md documenting the system for maintainers
• Initial 147KB AGENTS.md with complete Aurora DSQL documentation

The AGENTS.md file provides AI agents with a single, comprehensive resource covering setup, configuration, best practices, and troubleshooting for Aurora DSQL.

Issue #, if available:

Description of changes:

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.

[anwesham-lab]

workflow itself seems to be failing, I also think we're taking the wrong approach here.

rn there's a lot of content that agents don't need ie user onboarding, console instructions etc.

we want to keep in mind the agents.md files in discoverable repositories are supposed to help guide agents to optimal approaches efficiently as an alternative to skills. it should probably never exceed 500 lines realistically, maybe if we have a lot of directory trees.

it seems like agents-md compose simply concatenates all markdown files from the documentation/docs/**/*.md which ends up with this massive, hard to read, bloated AGENTS/md file.

• No curation - everything is included
• No summarization - full documents included verbatim
  • Redundant content - things like {% include 'copy-page-script.md' %} are left in

The solutions I can think of, seem to take away from automation though ie:

---

Instead of including everything, use a more selective config:

export default {
  include: [
    'documentation/docs/agents-md/**/*.md',  // Agent-specific content only
    '**/*.agents.md',
  ],
  exclude: [...],
} satisfies AgentsMdConfig

This requires creating dedicated *.agents.md files that are summaries/condensed versions.

Maybe you can get less manual work by adding some kind of Bedrock-driven summarization workflow with the our team's github AWS account? How do we do AWS work on other github repos? Should you use taht?

---

Explicitly optioning a summarization pipeline:

Add a build step that:

1. Reads full documentation
2. Uses an LLM or heuristics to summarize/condense each file
3. Outputs condensed versions to agents-md/ folder
4. agents-md compose then combines these condensed files

---

Create a template (will require manual updates over time) that:

• Extracts key sections from documentation (e.g., headers, code examples, important warnings)
• Generates a structured AGENTS.md with:
  • Quick reference
  • Key concepts
  • Common patterns
  • Links to full docs for details
• Reference detailed docs via URLs or file paths
• Let agents fetch details on-demand