fea: Introduce Lola package manager

[dmartinol]

Summary

Replace agent specific installation instructions with a marketplace for Lola.

• Renamed .mcp.json to mcps.json.
• Updated installation instructions in all the collections.
• Updated design specs, linter and design validators

Pack(s) affected

• rh-sre
• rh-developer
• ocp-admin
• rh-support-engineer
• rh-virt
• repo-wide

Change type

• New skill
• New agent
• New pack
• Update existing skill / agent
• MCP server config (.mcp.json)
• Docs / README
• CI / tooling

CLAUDE.md compliance

• Agents orchestrate skills; no direct MCP/tool calls in agents
• Skills are single-purpose task executors
• Skills encapsulate all tool access (MCP tools invoked only inside skills)
• Document consultation: file is read with the Read tool, then declared to the user
• No credentials hardcoded; env vars used via ${VAR} references
• Human-in-the-loop confirmation added for any destructive or critical operations

Validation

• make validate passes locally
• New/changed skills have valid YAML frontmatter (name, description)
• New/changed agents have valid YAML frontmatter (name, description)

[github-actions]

Gemini Code Review

1. Summary

This PR migrates the repository to use the Lola package manager format. It renames all .mcp.json files to mcps.json, updates documentation and validation scripts to reflect this change, removes Claude-specific plugin.json files, and introduces a new marketplace/rh-agentic-collection.yml registry file. It also updates the documentation site UI to display Lola installation instructions.

2. Project Rules Compliance

• ❌ Violation - Principle Rh developer/troubleshooting #8 (MCP Server Availability Verification): In rh-virt/skills/vm-clone/SKILL.md and rh-virt/skills/vm-create/SKILL.md, the Human Notification Protocol was condensed into a single line, removing the explicit instructions to "Stop immediately", "Request decision", and "Wait for user input". This violates the required pattern in SKILL_DESIGN_PRINCIPLES.md. The AI needs these explicit steps to halt execution and ask the user how to proceed.
  • Note: The same single-line violation already existed in vm-delete, vm-inventory, vm-lifecycle-manager, and vm-rebalance. Since you are modifying these lines to update the filename, please restore the proper multi-line format as defined in rh-virt/SKILL_TEMPLATE.md.
• ⚠️ Build reminder: Since skills and mcps.json files were changed, please remember to run make validate locally to ensure all structure and design principles pass.

3. Code Quality Issues

• Potential CI Failure (plugin.json): The PR deletes .claude-plugin/plugin.json from several packs. However, scripts/validate_structure.py likely still contains a validate_plugin_json() function. If the script isn't updated to treat plugin.json as optional (or removed entirely), make validate and the CI pipeline will fail. Please verify that validate_structure.py handles the absence of plugin.json gracefully.

4. Suggestions

• Missing Pack in Marketplace: rh-support-engineer was present in the old marketplace.json but is missing from the new marketplace/rh-agentic-collection.yml. I noticed it's also excluded from the docs generation (DOCS_EXCLUDED_PACKS in generate_pack_data.py), so this might be intentional. Just calling it out to ensure it wasn't accidentally dropped.

5. Verdict

REQUEST_CHANGES — The condensation of the Human Notification Protocol in the rh-virt skills violates the project's safety rules for prerequisite verification. Please restore the multi-line format that explicitly instructs the agent to stop and wait for user input.

---

Automated review by Gemini 3.1 Pro | Rules: CLAUDE.md, SKILL_DESIGN_PRINCIPLES.md

[r2dedios]

Lola installation works. Great Job! Just a few comments to check