Fix test_all_artifacts_have_required_fields silently skipping half th…

[mpilking]

Problem

test_all_artifacts_have_required_fields validates 27 of 53 artifacts. The other 26 are silently skipped.

The test derives a filename from each artifact's name field, then gives up when the lookup misses:

art = loader.get_artifact(name.lower().replace(" ", "_").replace("-", "_"))
if art is None:
    # Try the description-based name
    continue

That only resolves when an artifact's name directly derives to its filename, which often fails:

┌────────────────────┬────────────────────┬───────────────────┐
│        name        │      derives       │  actual filename  │
├────────────────────┼────────────────────┼───────────────────┤
│ Authentication Log │ authentication_log │ auth_log.yaml     │
├────────────────────┼────────────────────┼───────────────────┤
│ BASH History       │ bash_history       │ bash_history.yaml │
├────────────────────┼────────────────────┼───────────────────┤
│ Cron Jobs          │ cron_jobs          │ crontab.yaml      │
├────────────────────┼────────────────────┼───────────────────┤

Every miss hits the continue and the artifact goes unchecked with no signal. 22 Windows and 4 Linux artifacts are currently affected.

Fix

Read the artifact directories directly, matching the seven other _load_all_in_dir loops already in this file — this was the only artifact test not using that idiom.

Result

Coverage goes from 27 to all 53 artifacts, and the test passes unchanged: every artifact does carry name, description and platform. No artifact files are modified.

🤖 Generated with Claude Code (https://claude.com/claude-code)