fix(tests): purge deleted tests, bill their storage, and record versions as loaded

[TheodoreSpeaks]

Addresses the workflow-test review comments on the v0.9.17 release PR (#8853).

Changes

• Deleted tests block cleanup (Greptile P1): retention cleanup deleted a test's source file while the soft-deleted workflow_test row still referenced it, so the foreign key rejected the delete on every run. Cleanup now removes expired test rows (their runs cascade) in the same transaction, before the file.
• Deleted tests keep consuming quota (Greptile P1): test sources are billed on create, but cleanup treated every non-workspace context as unbilled. test files now go through the billed path, which releases their versions and decrements the payer.
• Changed workflows look tested (Greptile P1): the run recorded draft timestamps and child deployments after it finished. The executor now reports the version it loaded (currentWorkflow on its delegation origin) through enterWorkflow, and the session records each draft's updatedAt at that point.
• Org-chat submenus offered tests (Cubic): workspace submenus now exclude the composer's excluded types.
• Open test (Cubic): opens a new tab on web, like other internal links.
• Doc fixes: hook inheritance exceptions (custom-block sources, workflows called as Agent tools) and mockSampleOutput wording.

Verification

• New background/cleanup-soft-deletes.integration.ts against real Postgres: an expired test, its run, and its source are purged, a live one is kept, and workspace + payer storage drop by the source's bytes. It fails on the pre-fix cleanup.
• Real in-process test-file run of a parent → child workflow: recorded versions match the DB exactly for both workflows, in draft (updatedAt) and deployed (active deployment ids).
• lint:check, check:audits, and the executor / workflow-tests / cleanup unit suites pass.

Companion PRs

• simstudioai/mothership#665

🤖 Generated with Claude Code

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

1 Skipped Deployment

Project	Deployment	Actions	Updated
docs	Skipped		Oct 9, 2026 6:03pm UTC

[github-actions]

⚠️ Cross-repo companion check

One or more companion PRs aren't merged into staging yet. Merging this without them will leave copilot and sim out of sync — merge them in lockstep.

• ❌ simstudioai/mothership#665 — OPEN, not merged (targets staging) — fix(worker): describe mockSampleOutput as setting the mock's answer

[TheodoreSpeaks]

@greptile review

[cubic-dev-ai]

All reported issues were addressed across 13 files

<sub>Reply with feedback, questions, or to request a fix.

Turn on auto-fix | Re-trigger cubic</sub>

[greptile-apps]

Confidence Score: 5/5

[Critical impact] No new blocking issue was found in the changes since the previous review.

Summary

This PR fixes deleted-test cleanup and storage accounting, records loaded workflow versions, and adjusts test links and composer menus. The latest revision reads draft timestamps before starting the executor.

• Expired tests and their sources are removed together, freeing their storage.
• Test results name the workflow versions that actually ran.
• Organization chat menus leave excluded resources out of workspace submenus.
• Embedded tests open in a new web tab and stay in place on desktop.
• Test help describes sample outputs as mock answers.

Diagram

%%{init: {'theme': 'neutral'}}%%
flowchart TD
  A[Start a draft test] --> B[Read workspace draft timestamps]
  B --> C[Load and execute workflow]
  C --> D[Record entered workflow versions]
  D --> E[Save test result and execution links]
  E --> F[Compare saved timestamps with current drafts]

Loading

_{Reviews (2) · Last reviewed commit: "fix(tests): take a draft's timestamp fro..." · Reviewed by Greptile}

[TheodoreSpeaks]

@greptile review