Flaky Test Investigator

flaky-test-investigator

Purpose

Analyze why an automated test is unstable and propose practical fixes around timing, ordering, environment, data, or assertions.

Trigger this skill when

• You need automation work at this specific test level or concern rather than a vague “write some tests” request.
• You want explicit artifacts, assertions, and maintenance tradeoffs instead of brittle generated code.
• You need a reusable output that can hand off cleanly to the next testing skill.

Expected inputs

• failing test behavior
• execution logs
• timing symptoms
• environment details

Deliverables

• flake hypotheses
• diagnostic plan
• stabilization actions
• residual risk notes

Operating procedure

1. Clarify the target behavior, boundary, or automation goal and what would count as evidence.
2. Structure the test asset so setup, action, and assertions are observable and separated.
3. Minimize hidden assumptions around data, timing, environment, and ordering.
4. Produce lean outputs with enough rationale that another agent or developer can continue the work safely.

Quality gates

• Prefer stable, explainable automation over superficially broad coverage.
• Keep assertions tied to observable outcomes, not vague expectations.
• Be explicit about flake risks, environmental dependencies, and setup or cleanup needs.
• Do not hide uncertainty; name assumptions and residual risk.

Handoff targets

• test-isolation-reviewer
• webdriver-flow-writer
• regression-suite-curator

Output style

• Be explicit about uncertainty, infeasibility, and residual risk.
• Prefer compact, explainable artifacts over bloated lists.
• Tie tests back to behaviors, contracts, or flows.
• Make expected outcomes observable.

Failure modes to avoid

• Do not over-automate unstable or poorly specified behavior without naming the risk.
• Do not create tests whose assertions are weaker than the stated goal.
• Do not bury fixture, selector, timing, or dependency assumptions.

Minimum output skeleton

## Summary
## Findings
## Structured outputs
## Coverage / rationale
## Assumptions
## Open questions
## Recommended next skill