monitor-cilisted

# Monitor CI CI is the final gate before review, and every failure must be resolved. It is also long-running and noisy, and unstructured polling burns context, so this skill fixes which commands to run, how often, and how to act on each failure type. The loop is **observe → classify → fix → re-push → re-observe**, with a retry budget. When the budget runs out, stop and escalate. --- ## CI Job Map (typical Agora service repo) The `main` workflow on Agora service repos under `a-novel/service-*` runs on every push to any branch. The job set varies by repo; the table below is the common surface across those services. For the jobs on the current checkout, run `gh pr checks <n>` (or read `.github/workflows/main.yaml`) and intersect with the rows below; anything unlisted is repo-specific or downstream of a base table entry. | CI Job | What it checks | Local equivalent | Typical failure | | ----------------------------- | ------------------------------------------ | ----------------------------- | ---------------------------------------------------------------------------- | | `generated-go` | `go generate ./...` is up to date | `pnpm generate:go` | Forgot to run `pnpm generate:go` after proto/interface | | `lint-go` | `golangci-lint run` clean | `pnpm lint:go`