| name | copilot-schema-pricing-update |
| description | Run Copilot Usage Studio's weekly maintenance workflow after a VS Code or GitHub Copilot update, especially when checking VS Code Agent Debug Log schema drift, new Copilot model ids, GitHub Copilot pricing changes, pricing docs, schema baselines, and compatibility gates. |
Copilot Schema And Pricing Update
Use this skill when maintaining Copilot Usage Studio after a VS Code, GitHub Copilot Chat, or GitHub Copilot pricing update.
Workflow
- Read project context first:
AGENTS.md
NEXT.md
docs/schema-change-workflow.md
docs/pricing.md
docs/debug-log-schema.md
- Check the working tree with
git status --short. Preserve unrelated user changes.
- Run the schema audit:
npm run schema:audit
- Review
tmp/schema-audit.md.
- Treat
breaking as a blocker.
- Treat
warning as a probe-coverage or compatibility question; do not accept the baseline from one narrow session.
- Treat additive model ids, event types, and fields as feature candidates only after their meaning is understood.
- Check GitHub's official Copilot pricing page:
https://docs.github.com/en/copilot/reference/copilot-billing/models-and-pricing
- Update
data/github-copilot-pricing.json only from the official table.
- Update
version, snapshotDate, and importedAt.
- Preserve normal input, cached input, cache write, output, and long-context tiers.
- Add raw model-id aliases when VS Code logs ids that differ from GitHub's display label.
- Update consumers and docs when pricing changes:
scripts/pricing-utils.mjs
src/app/pricing.ts
- relevant tests under
scripts/*.test.mjs or src/app/*.spec.ts
docs/pricing.md
docs/data-ingestion.md
docs/debug-log-schema.md
CHANGELOG.md
- Run compatibility gates:
npm run test:scripts
npm test -- --watch=false
npm run build
npm run vscode:package when the change can affect the extension.
Baseline Rule
Do not run npm run schema:accept just because the latest audit completed. Accept the baseline only after controlled probes cover the behavior in docs/schema-change-workflow.md, warnings are understood, and any scanner/docs/test changes are complete.
Evidence Rules
- Prefer
copilotUsageNanoAiu source usage when present.
- Keep token-bucket pricing as the explanation and fallback layer.
- Do not use
models.json catalogue token-price metadata as the authoritative price table.
- Do not claim missing fields are removed unless a targeted probe should have produced them.
- Do not commit raw prompts, tool arguments, file paths from local logs, generated
sessions.json, or ignored tmp reports.