feat: tooling API AI skill

[chris-de-leon-cll]

core ref: 2a52a992df57d88d85e2d6d6aad393950af0944a

Summary

Adds a repo-local AI skill for writing and reviewing CCIP deployment tooling code based on the AI Skill Guidelines here.

The skill lives at .agents/tooling/SKILL.md and gives agents a focused workflow for changesets, adapters, registries, datastore flows, integration tests, and other operator-facing automation.

The goal is to make future AI-assisted tooling work more consistent with this repo's deployment patterns without turning the skill into generic engineering advice.

What Changed

• Added .agents/tooling/SKILL.md.
• The skill instructs agents to consult:
  • deployment/docs/style-guide.md for changeset authoring guidance.
  • deployment/docs/style-guide-contributions.md when updating the style guide itself.
• The skill captures practical rules for tooling code:
  • start from nearby implementations and existing extension points,
  • make operators provide intent rather than fragile internal state,
  • resolve refs before using address/version/type fields,
  • use the narrowest clear abstraction,
  • keep apply paths retry-safe,
  • validate only the next meaningful assumption,
  • test operator-facing behavior instead of abstraction mechanics.

Reviewer Notes

This is intentionally a single lean SKILL.md, with no extra reference files. The deployment style docs are linked rather than copied so the skill stays small and the docs remain the source of truth.

[github-actions]

Metric	feat/tooling-ai-skill	main
Coverage	70.1%	69.8%