Add annual-cloud-data-update skill (yearly metadata refresh workflow)

[adrianco]

Summary

Adds a Claude Code skill under .claude/skills/annual-cloud-data-update/ that documents and drives the yearly refresh of Cloud_Region_Metadata.csv when AWS, Google, Azure and Oracle publish their sustainability data. It captures the institutional knowledge for doing this cleanly so future updates are repeatable rather than rediscovered each year.

Three-phase workflow it encodes

1. Extract each vendor's metrics and their verbatim metric definitions into metrics/<year>/<vendor>-<year>-metrics.md — including downloading AWS's ~32 per-region fact sheets and Microsoft's detailed Environmental Data Fact Sheet, so the definitions (not just the numbers) are recorded.
2. Merge only metrics whose definition genuinely fits the schema column — additively, preserving CRLF, never modifying an already-published year.
3. Gaps report — metrics/<year>/gaps-<year>.md, a cross-vendor alignment analysis of what each vendor should change for comparable metrics (serves the project's lobbying mission).

Key rules it bakes in

• Per-region physical data only, not the global market claim — don't populate the CFE columns with a uniform "100% renewable" market claim; AWS's per-grid renewable list is per-region and OK.
• Fiscal-vs-calendar timescale — Azure (Jul–Jun) and Oracle (Jun–May) fiscal years end mid-calendar-year, so their "2025" isn't a calendar-2025 window; record exact periods and flag in the gaps report.
• The practical traps: CRLF line endings, awk mis-splitting the quoted geolocation comma, the unreliable AWS scraper, region-carbon-info lagging Google's PDF, and the additive-only (removed: 0) verification.

Structure

.claude/skills/annual-cloud-data-update/
├── SKILL.md
└── references/{aws,google,azure,oracle}.md
    + metrics-doc-template.md, gaps-report-template.md

Contains no data changes — purely the process documentation. Pairs with the 2025 data PR (#162) and the Oracle PR (#163).

🤖 Generated with Claude Code