Add experimental agent path-to-production playbook

[James Nguyen (james-tn)]

Add experimental agent path-to-production playbook

Asset

• Asset id: assets/agent-path-to-production
• Kind: playbook
• New asset, version 0.1.0, maturity experimental.

Summary

Add an evidence-led delivery guide for taking an agentic application from early
design or a pilot to an operable production workload. It includes six-lens
triage, specialist routing, three bounded delivery plays, release/operation
gates, and a blank private-workload readiness-record template.

The contribution is generic guidance, not a deployable accelerator, certification
of readiness, or claim of field validation. Organization-specific service
catalogues and engagement records are deliberately excluded.

Pillar chapters

docs/pillars/ adds an index and one chapter per lens:

1. AI CoE & Governance
2. AI Landing Zones
3. Agent Ops
4. AI Security & Observability
5. Solution Optimization
6. Proven Patterns for Agent Architectures

Each chapter uses the same eight sections, in the same order:

1. Pillar definition
2. Field signals
3. Entry questions
4. Approach and play
5. Readiness evidence and gate
6. Common pitfalls
7. Collaboration boundaries
8. References

How the chapters connect to the rest of the asset:

• The approach steps link to runbook Plays A to C.
• The gate tables use the same gate names as the readiness record.
• The runbook triage table and the README lens table now link to each chapter.
• Two lens names now match the chapter titles: "AgentOps" is now "Agent Ops",
  and "Agent Architecture Patterns" is now "Proven Patterns for Agent
  Architectures".

Content provenance

The field signals describe common situations in general terms. They are not
case studies. The chapters contain no customer, industry, product-of-customer,
person, or program names, no internal links, and no numeric outcome claims.
References are limited to public Microsoft Learn pages: Cloud Adoption
Framework, Well-Architected Framework, Azure Architecture Center, and Microsoft
Foundry documentation.

Validation

• python tools/assets.py validate: passes, with 0 errors.
• python tools/assets.py catalog and python tools/assets.py catalog --check:
  the catalog is up to date.
• git diff --check and git diff --cached --check: no whitespace errors.
• The repository has no link checker, so links were checked with a script:
  • Every relative link and heading anchor in the asset resolves.
  • Every new Microsoft Learn reference returns HTTP 200.
• A public-safety scan found no internal names, internal hosts, placeholders,
  or numeric duration claims.

The manifest schema does not list individual files, so manifest.yaml is
unchanged. The version stays at 0.1.0 because the asset has not been
published, and the changelog entry for that version now covers the chapters.

Checklist

• One asset with a manifest, README and changelog.
• python tools/assets.py validate passes.
• Generated catalog is current.
• No customer data, internal links, program assignments or secrets.
• Examples are generic; no third-party content is reproduced.
• New asset has an initial version and review date.

Curator technical and public-safety review remains required before merge.