Typical setups
In brief: Common QA automations map clear triggers to catalog workflows—fix high-severity issues, implement ready stories, repair failed batches, and gate risky runs with human approval.
Use these as starting points in the create wizard. Adjust conditions, policies, and invocation strategy to your team’s risk tolerance.
Auto-fix high-severity issues
Goal: When a serious bug lands (and it was not filed by ExploreChimp noise), start fix-issue.
| Field | Suggested value |
|---|---|
| Entity / event | Issue → Created (or Severity updated) |
| Conditions | severity = HIGH (or Critical); optionally reported_by ≠ __explorechimp__ |
| Workflow | fix-issue |
| Task template | Fix issue {{issue.id}} — {{issue.title}} |
| Aggregation | Immediate or Buffered with short grace if reporters edit severity quickly |
| Invocation | OpenHands or GitHub Issue |
| Human gates | Invoke approval on until you trust the agent; then off for P1 overnight coverage |
Implement when a story becomes ready
Goal: Development agent picks up implementation as soon as planning marks the story ready.
| Field | Suggested value |
|---|---|
| Entity / event | Story → Status updated |
| Conditions | status = ready |
| Workflow | implement |
| Task template | Implement story {{story.id}} — {{story.title}} |
| Aggregation | Buffered (grace ≥ 5 min) so last-minute edits settle |
| Human gates | Require approval before plan execute so humans review the plan PR / plan file |
Mirror the same pattern for Scenario → status ready when you implement at scenario granularity.
Repair failing CI / automation batches
Goal: After a SmartTest batch completes with failures, investigate and fix.
| Field | Suggested value |
|---|---|
| Entity / event | Test execution batch → Completed |
| Conditions | failed_count > 0 |
| Workflow | fix-test-execution or upkeep |
| Task template | Reference {{batch.batch_id}}, {{batch.branch_name}}, failed counts |
| Aggregation | Buffered so one push’s retries collapse into a single agent run |
| Invocation | GitHub Issue often fits CI-centric agent routing |
For a single flaky test rather than the whole batch, trigger on Test execution → Failed with Immediate aggregation and a template using {{test_execution.job_id}} / file path.
Run QA when a release moves to Ready
Goal: Kick a composite Run QA (or release checks) as the release lifecycle advances.
| Field | Suggested value |
|---|---|
| Entity / event | Release → Status updated |
| Conditions | lifecycle / status = Ready (match your release status vocabulary) |
| Workflow | run-qa or run-release-check |
| Task template | Include {{release.release_version}}, branch, prior version |
| Aggregation | Immediate (releases change infrequently) |
| Human gates | Invoke approval recommended for full composite runs |
Upkeep digest on a schedule window
Goal: Periodically close coverage gaps and hygiene work without firing on every small event.
| Field | Suggested value |
|---|---|
| Entity / event | Scenario or Story → Status updated (or batch completed) |
| Conditions | Narrow to ready / failed as needed |
| Workflow | upkeep |
| Aggregation | Windowed (e.g. every 1–6 hours) |
| Task template | List aggregated identities; policy upkeep.policy.md or a weekly variant |
Requirement quality when a draft is promoted
Goal: Gate agent coding on clean specs via run-requirement-quality-checks.
| Field | Suggested value |
|---|---|
| Entity / event | Story → Status updated |
| Conditions | Previous draft → current ready (or status = ready) |
| Workflow | run-requirement-quality-checks |
| Aggregation | Buffered |
| Human gates | Optional; failures create follow-up work before implement automations fire |
Pair with a separate implement automation that also requires ready and (by process) a green quality check, or keep implement behind plan-execute approval.
ExploreChimp-driven issues (usually exclude)
ExploreChimp can create many exploratory issues. For auto-fix automations, add:
reported_by ≠ __explorechimp__
…or route ExploreChimp findings to a triage automation with invoke approval instead of autonomous fix.
Choosing gates and strategies quickly
| Risk | Aggregation | Gates | Strategy |
|---|---|---|---|
| Low (label tidy, quality check) | Immediate / Buffered | None | GitHub Issue or OpenHands |
| Medium (fix single issue / test) | Buffered | Optional invoke approval | Either |
| High (implement, full run-qa) | Buffered | Plan-execute and/or invoke approval | Prefer the strategy your team already reviews PRs with |
Related
FAQ
How do I auto-fix high-severity issues?
Trigger on Issue created (or severity updated) with severity = HIGH, exclude ExploreChimp reporters if needed, and run the fix-issue workflow via OpenHands or GitHub Issue.
How do I react to failing CI batches?
Trigger on Test execution batch completed with failed_count > 0, use buffered aggregation, and run fix-test-execution or upkeep.
Should implement automations require approval?
Yes for most teams—use require approval before plan execute so humans review the plan before the agent continues.