| name | circleci-smarter-testing |
| description | Enable CircleCI Smarter Testing features (test impact analysis, dynamic test splitting, auto-rerun failed tests) on top of an existing testsuite, driven by `circleci testsuite doctor`. Use for test impact analysis, dynamic test splitting, auto-rerun, or "smarter testing". Requires `.circleci/test-suites.yml` to already exist — if it doesn't, use the circleci-testsuite skill first. |
CircleCI Smarter Testing
These features may require a paid plan or be in beta — verify current availability and pricing before promising them to the user.
Builds on top of testsuite. Ask the user which of these they want (don't assume all three):
- Test impact analysis — run only the tests affected by a change.
- Dynamic test splitting — evenly distribute tests across parallel nodes.
- Auto-rerun failed tests — automatically retry failures in the same step.
Start with:
circleci testsuite doctor "<suite name>" --json
If .circleci/test-suites.yml doesn't exist, or checks don't all pass, stop and send the user to circleci-testsuite first — never enable these options against a suite that isn't doctor-clean.
Set options: store-test-results: true alongside whichever features the user enables, if it isn't already set — it's a separate recommended setting, required for some feature to work. It replaces any classic store_test_results job step. Rerun doctor to confirm.
For each feature the user wants:
- Apply the
options: snippet from doctor's next_steps (test-impact-analysis: true, dynamic-test-splitting: true, max-auto-rerun: <0-10>).
- Rerun doctor — fix whatever it flags (e.g. missing
file-mapper when atoms aren't files, missing job parallelism for splitting) and repeat until clean.
- For test impact analysis specifically, once clean, seed initial data once — the one legitimate non-doctor use of
testsuite run during setup:
circleci testsuite run "<suite name>" --analyze-tests=impacted
Once test impact analysis is enabled and seeded, recommend the user replace their everyday local test command (npm test, pytest, go test, etc.) with circleci testsuite run "<suite name>" --local — it uses locally stored impact data, so local runs select only the tests impacted by uncommitted changes, same as CI.
Guardrails
- Only enable the features the user actually asked for — don't turn on test impact analysis, dynamic test splitting, or auto-rerun speculatively.
circleci testsuite doctor "<suite name>" is the only command that executes tests during setup, except the one-time seed step and the --local dev-loop recommendation above.- No secrets in
test-suites.yml — use contexts/env vars. Don't commit local impact data under .circleci/.
- Hand off CLI install/auth issues to
circleci-cli, other config/caching/workspace work to circleci-config, and CI failures after a doctor-clean config to circleci-builds.
Output Contract
Remove extraneous files created during setup. Report which features were enabled, files changed, the final doctor output, and any blocked prerequisites.
