Installs into .claude/skills of the current project.
Are you the author of Intuitive Tests?
Add the live security badge to your README. It updates with every re-scan.
[](https://www.skillsdirectory.com/skills/miaodx-intuitive-tests)
---
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. Keep at least one
meaningful proof of every 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](references/value-and-pruning.md) for test admission or
deletion decisions; read [suite structure](references/suite-structure.md) 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.