Unified Test Runner

Run all test suites sequentially. Both system tests leave their output on disk so Step 4 can cross-validate the shared SQLite store against all markdown files, then Step 5 cleans up.

Flags

Parse the skill args for these flags:

Steps

Step 1: Unit + Integration Tests

Skip if --skip-unit was passed.

Run: go test ./... -count=1

• •Pass criteria: exit code 0, all tests pass.
• •Print: Step 1: Unit tests — X packages, PASS/FAIL

Step 2: Single Data Source System Test

Skip if --skip-system was passed.

Invoke /test-single-datasource-db --no-cleanup, also passing --verbose if present.

The --no-cleanup is always passed here regardless of user flags — we need the files on disk for Step 4.

• •Pass criteria: Skill reports ALL TESTS PASSED.
• •Print: Step 2: Single datasource — PASS/FAIL

Step 3: Double Data Source System Test

Skip if --skip-system was passed.

Invoke /test-double-datasource-db --no-cleanup, also passing --verbose if present.

Same as Step 2 — --no-cleanup always passed to keep files for Step 4.

• •Pass criteria: Skill reports ALL TESTS PASSED.
• •Print: Step 3: Double datasource — PASS/FAIL

Step 4: Cross-Integration — SQLite ↔ Markdown Consistency

Skip if --skip-system was passed.

This validates that the shared _notion_sync.db at test-output/ is consistent with all the markdown files produced by both system tests.

Run these checks using sqlite3 (read-only) and filesystem inspection:

4a. SQLite has pages from both databases

SELECT database_id, COUNT(*) FROM pages WHERE deleted = 0 GROUP BY database_id;

• •Pass criteria: Two distinct database_id values are present:
  • •2fe57008-e885-8003-b1f3-cc05981dc6b0 (single-source) with 11 pages
  • •c9aa5ab2-b470-429c-ba9c-86c853782bb2 (double-source) with >= 13 pages

4b. Every non-deleted SQLite page has a matching .md file

For each row in pages where deleted = 0 and file_path is not empty:

• •

  Verify the file at file_path exists on disk.

• •

  Pass criteria: All file paths resolve to existing files.

4c. Every .md file with notion-id has a matching SQLite row

Scan all .md files under test-output/ for notion-id in frontmatter. For each:

• •

  Query SELECT id FROM pages WHERE id = '<notion-id>' AND deleted = 0

• •

  Verify a row exists.

• •

  Pass criteria: All markdown notion-ids found in SQLite.

4d. Timestamps match between SQLite and frontmatter

For a sample of 3 pages (pick one from single-source, two from double-source):

• •

  Compare last_edited_time in SQLite vs notion-last-edited in the .md frontmatter.

• •

  Pass criteria: Timestamps match (use timestampsEqual logic — .000Z vs Z is OK).

4e. FTS index covers both databases

SELECT COUNT(*) FROM pages_fts;

• •Pass criteria: Count >= total non-deleted pages across both databases.

Print: Step 4: Cross-integration SQLite consistency — PASS/FAIL

Step 5: Cleanup

Skip if --no-cleanup was passed. Print Step 5: Skipped (--no-cleanup) and leave files.

Otherwise, clean up everything:

1. •Delete test-output/test database obsdiain complex/ (single-source folder)
2. •Delete test-output/test database - double data source/ (double-source folder)
3. •Clean SQLite: delete rows from pages where database_id IN the two test database IDs. Use Python or Go to execute the SQL.
4. •If _notion_sync.db has zero rows remaining in pages, delete the .db file entirely.
5. •If test-output/ is now empty, delete it.

Print: Step 5: Cleanup — done

Summary

Print a combined results table:

| Suite                          | Result  |
|--------------------------------|---------|
| Unit tests (go test)           | PASS    |
| Single datasource system       | PASS    |
| Double datasource system       | PASS    |
| Cross-integration (SQLite ↔ MD)| PASS    |

Then:

• •If all suites passed: ✅ ALL SUITES PASSED
• •If any failed: ❌ SOME SUITES FAILED followed by which ones failed