diff --git a/AGENTS.md b/AGENTS.md index f03d354be39..9aceb79a189 100644 --- a/AGENTS.md +++ b/AGENTS.md @@ -4,10 +4,10 @@ - **Test**: `pnpm test [path/to/file.test.ts] --maxWorkers=2`; incremental: `pnpm test:changed [path] --maxWorkers=2` ([details](.github/CONTRIBUTING.md#targeting-local-tests)). Separate rendered UI E2E checks need explicit, reproducible viewport/state. - **Lint/format**: Basic: `pnpm exec oxlint `; full lint gate: `pnpm lint`; formatting: `pnpm exec oxfmt --check ` (fix: `--write`). Applicable gates include `pnpm check:code-quality:changed ` and `pnpm check:flaky-test-pattern-ratchet`. -Optional: `pnpm check:changed --maxWorkers=2`; if selection fans out excessively, use scoped checks and report limits. Use the PR base (default `origin/main`; `ORCA_CODE_QUALITY_BASE=` for stacks). `--base=HEAD` and dirty-only checks provide incremental evidence only. - > [!IMPORTANT] -> Always lint changed files before making/pushing-to PRs (so CI never fails on it) +> Run `pnpm check:changed` after your last edit/rebase, right before every push. Never `--no-verify`. + +If selection fans out excessively, use scoped checks and report limits. Use the PR base (default `origin/main`; `ORCA_CODE_QUALITY_BASE=` for stacks). `--base=HEAD` and dirty-only checks provide incremental evidence only. # Don't Poll the CI