Skill v1.0.2
currentAutomated scan100/1007 files
version: "1.0.2" name: intuitive-tests description: Audit or improve test-suite structure and signal when the user asks about redundancy, pruning, layout, fixtures, or parameterization. disable-model-invocation: true
Intuitive Tests
Improve test-suite signal and maintainability. Keep proof of observable behavior, real failure modes, and supported contracts; consolidate duplicates and remove assertions that merely restate implementation shape. Do not delete the last meaningful proof of a behavior. Runtime discovery and packaging checks can be valuable contracts even when they look structural.
For broad or ambiguous cleanup, audit then propose a decision-complete slice. For a precise implementation request, execute that slice and park unrelated ideas. Keep existing repo test conventions; pytest examples are not a required taxonomy.
Modes
| Mode | Use when | Output | Redirect when | |
|---|---|---|---|---|
| Audit / propose | Cleanup target is unclear | Evidence, candidates, bounded recommendation and proof plan | A slice is already approved | |
| Prune / consolidate | Tests duplicate proof or bind internals | Smaller suite with behavior coverage preserved | No redundant proof exists | |
| Marker / layout | Suite selection or paths need organization | Updated runners, consumers and collection proof | The issue is test value | |
| Fixture / parameterize | Repeated setup or cases obscure intent | Shared setup or cases with readable diagnostics | Reuse is speculative |
Read value and pruning for test admission or deletion decisions; read suite structure only for markers, moves, fixtures, or parameterization. An audit may recommend a route without loading every implementation procedure.
Verify the changed behavior and affected runner/collection paths. Derive any external-service, hardware, or credential limits from current user/repo policy. Report kept/merged/deleted/reclassified tests, remaining proof, affected commands, and skipped checks with their actual reason.