tests/universal-test-harness.md

# Universal Test Harness

Purpose: define the three validation checks every Universal Capability Object (UCA) must pass before it can move from `proposed` to `approved` in the Skill Workshop Queue.

## 1. Schema test

Validate the UCA instance against `schemas/universal-skill.schema.json` using a standard JSON Schema validator.

Pass criteria:
- all `required` fields present
- `id` matches `^[a-z0-9]+(-[a-z0-9]+)*$`
- `category` and `status` are valid enum values
- at least one entry in `adapters.providers` or `adapters.runtimes`

## 2. Dry-run test

Walk through `steps` without executing any real side effects (no filesystem writes, no network calls, no commands). Confirm:
- each step has a clear input/output
- no step depends on a `context_needed` item that isn't declared
- the final step produces every declared item in `outputs`

## 3. Real usage example

Pick one realistic end-user request that should trigger this skill. Run it through the full pipeline:

```
User goal → Capability Detector → Registry Radar → Capability Classifier
→ Trust Scanner → Adapter Generator → Test Runner → Skill Workshop Queue
```

Pass criteria:
- Capability Detector produces the expected `capability intent`
- Registry Radar correctly reports match/no-match against `registry/skills.index.json`
- Adapter Generator produces one adapter file per requested target, each conforming to the format described in `adapters/provider-adapter-map.md` or `adapters/runtime-adapter-map.md`

## Result states

| Result | Next queue state |
|---|---|
| All 3 checks pass | `tested` → eligible for `approved` |
| Schema test fails | stays `proposed`, returned to Adapter Generator for fixes |
| Dry-run fails | stays `inspected`, steps need rewriting |
| Real usage example fails | stays `tested`, logged to Luuna Memory as a self-improvement note (see `agents/luuna-skill-forge.agent.json` → `self_improving_loop`) |

## Minimal test log format

```json
{
  "skill_id": "example-skill",
  "version": "0.1.0",
  "run_date": "YYYY-MM-DD",
  "schema_test": "pass | fail",
  "dry_run_test": "pass | fail",
  "real_usage_example": "pass | fail",
  "notes": ""
}
```