Unpublish finishes in the request, but it returned a job ID that
`get-publish-job` could never find. It now returns the domains it
unpublished. A missing publish job also stays a normal not-found error
instead of suggesting a CLI update.
Closes#6578
## Changes
- Keep filename-based E2E shard discovery. Rebalance the files across
ten shards using measured CI times; each shard keeps its own test
environment and the same 12-minute safety timeout.
- Remove the fabricated unpublish job ID. `get-publish-job` remains for
publish jobs.
- Remove publish-only flags from the unpublish CLI and clarify CLI/MCP
guidance.
- Keep legacy optional server inputs for compatibility. The API input
contract does not change; clients that read the old `jobId` result need
to use `unpublishedDomains` instead.
- No database migration. Existing domain-removal behavior is unchanged,
including retention of other published domains.
## Checklist
- [x] Reproduced both reported failures with tests before the fix.
- [x] Reviewed the change for compatibility, redundant code, and
retained-domain behavior.
- [x] Updated the authoritative MCP docs and regenerated the CLI/MCP
docs.
- [x] Ran focused Builder, CLI, domain, and build tests; workspace
typecheck, lint, package-boundary, generated API, and generated-doc
checks.
- [x] Restore filename-based shard discovery and rebalance the files
across ten shards.
- [x] Verify every Chromium workflow test runs in exactly one shard and
update the E2E contributor docs. All 141 tests are assigned once across
the ten shards.
- [x] Verify parallel-safe long E2E files and the two-submission legacy
form test in CI.
- [x] Let independent slow E2E files run their tests concurrently; all
ten shard jobs passed.
- [ ] Confirm all E2E jobs finish in under five minutes. Eight did;
shard 1 took 5:10 and shard 4 took 6:32. Move the long Content-mode page
and shared-project bridge files to separate shards, then remeasure.
- [x] Required CI checks pass.
No live unpublish was run because it would change a published site.
Agent evaluations were not run; they require separate approval.
## What changed
- Return structured validation issues with stable codes, complete field
paths, constraints, and minimal correction examples.
- Preserve validation details through the Builder API, CLI, MCP, local
project sessions, and tRPC errors.
- Explain invalid expressions, JSX fragments, page selectors, component
inputs, asset updates, and project settings in terms an agent can act
on.
- Remove stack frames and credentials from parser details before
exposing them.
- Keep concise human-readable errors while making the complete
structured issues available to agents.
## Why
Agents previously received vague errors without enough information to
repair a request. They could repeat the same invalid call or require
manual intervention.
## User impact
CLI and MCP agents can now identify the exact invalid field, understand
the constraint, see a valid example, and retry with a corrected request.
## Validation
- Focused runtime, MCP, CLI, API, and tRPC validation tests.
- Structured error propagation tests across local and server execution.
- Typechecks and generated contract verification.
Oleg Isonen