Make workflow frontmatter visible to agents for simpler tool discovery

[Copilot]

Agents weren't discovering project-related safe outputs (update-project, create-project-status-update) because the workflow frontmatter was stripped during runtime-import.

Changes

• Frontmatter preservation: Modified runtime_import.cjs to preserve frontmatter instead of stripping it
• Visibility wrapper: Frontmatter is wrapped in <workflow-configuration> tags with YAML code blocks for agent visibility
• Test updates: Updated 3 tests to expect frontmatter preservation instead of removal
• Simpler approach: Removed complex Go-based categorization logic in favor of letting agents see the actual YAML configuration

Before/After

Before:

---
safe-outputs:
  update-project:
    project: "https://github.com/orgs/my-org/projects/1"
---
# Instructions
(frontmatter stripped, agent doesn't see configuration)

After:

<workflow-configuration>
```yaml
safe-outputs:
  update-project:
    project: "https://github.com/orgs/my-org/projects/1"

Instructions

(agent can see the configuration directly)

This simpler approach puts control in workflow authors' hands - they can see their own configuration and write instructions accordingly, rather than relying on automated categorization that may not match their intent.

<!-- START COPILOT CODING AGENT TIPS -->
---

💡 You can make Copilot smarter by setting up custom instructions, customizing its development environment and configuring Model Context Protocol (MCP) servers. Learn more [Copilot coding agent tips](https://gh.io/copilot-coding-agent-tips) in the docs.

[mnkiefer]

@copilot This is overkill. It would be enough if the instructions of workflow markdown file itself could see/know about the frontmatter above it.

[Copilot]

@copilot This is overkill. It would be enough if the instructions of workflow markdown file itself could see/know about the frontmatter above it.

You're right - I've reverted the complex categorization. Instead, I'll make the frontmatter visible to the agent by not stripping it during runtime-import, so the agent can see what safe-outputs are configured directly from the YAML. This is much simpler and puts control in the workflow author's hands.