fix(uipath-platform): run queue/trigger seed e2e under docker, not tempdir

[alexenica]

What & why

skill-platform-queue-progress-e2e and skill-platform-trigger-api-e2e fail their pre-run (seed.py) every run with:

No compatible version of '@uipath/orchestrator-tool' found for the CLI's 1.197.x line on npm or GitHub Packages.

Root cause is the sandbox.driver, not the seed. Both tasks set driver: tempdir, which runs on a bare host with no pre-baked tool plugins. seed.py calls uip or folders, so the CLI tries to auto-install @uipath/orchestrator-tool at runtime — which fails on the prerelease CLI the eval installs (it resolves tools on stable, and there's no stable build on its prerelease-only line; see UiPath/cli#3083). The pre-run dies before the task starts → score 0, deterministically.

skill-platform-role-lifecycle-e2e runs the same seed.py but under the default docker driver, where orchestrator-tool is pre-baked — and it passes. That's the tell.

The change

Switch both tasks to driver: docker, so the pre-baked skills-image:latest provides orchestrator-tool and the seed's uip or calls resolve with no runtime install. This matches the existing convention — the uipath-ixp e2e tasks already use driver: docker for exactly this reason (uip ixp is pre-baked, not resolvable in tempdir).

Why this over the CLI fix

UiPath/cli#3083 fixes the underlying prerelease resolution bug, but it only reaches these tasks after it merges and a fixed build is published and the eval installs it. This change fixes the two seed failures directly, now, with no cross-repo dependency. (#3083 is still worth landing as a real product bug.)

Verified by inspecting the 2026-07-15 run: of 778 tasks, only these two hit the error — the only two that both (a) run seed.py's uip or in pre-run and (b) use driver: tempdir.

🤖 Generated with Claude Code

[github-actions]

Claude finished @alexenica's task in 1m 58s —— View job

---

PR Review: fix(uipath-platform): run queue/trigger seed e2e under docker, not tempdir

• Gather context and read review criteria
• Read full diff and changed files
• Read surrounding context (sibling tasks, seed.py)
• Perform review and post findings

---

Summary

Switches driver: tempdir → driver: docker in two e2e task YAMLs (queue_progress_e2e.yaml, trigger_api_e2e.yaml) whose seed.py pre-run calls uip or folders, requiring @uipath/orchestrator-tool to be pre-baked. The tempdir driver lacks this, causing deterministic seed failures on prerelease CLI builds. The fix aligns with the existing convention — role_lifecycle_e2e (same seed.py) already uses the default docker driver and passes.

Change-by-Change Review

1. tests/tasks/uipath-platform/resources/queue_progress_e2e.yaml

Severity: OK

The change from driver: tempdir to driver: docker is correct. Verified that seed.py calls uip or folders get/create (lines calling uip_json("or", "folders", ...)) which requires orchestrator-tool. The 5-line comment block above sandbox: explains the rationale clearly and references the upstream CLI bug (UiPath/cli#3083). No other fields changed — tags, success_criteria, pre_run, and post_run are untouched.

2. tests/tasks/uipath-platform/resources/trigger_api_e2e.yaml

Severity: OK

Identical driver change and identical comment block. Same rationale applies — this task uses the same seed.py with the same uip or folders calls.

What's Missing

Nothing identified. The change is minimal, targeted, and consistent with the existing convention across sibling e2e tasks.

Area Ratings

Area	Status	Notes
Frontmatter	OK	No frontmatter changes; existing fields valid
E2E Tests	OK	Driver fix restores these two tasks to runnable state
Skill Body	N/A	No skill body changes
References & Assets	N/A	No reference/asset changes
Repo Hygiene	OK	No secrets, no unrelated changes, scoped to the two affected files

Issues for Manual Review

None found.

Conclusion

Clean, well-scoped fix. The root cause analysis in the PR description is thorough — confirmed by checking that seed.py calls uip_json("or", "folders", ...) and that the sibling role_lifecycle_e2e (which runs the same seed under the default docker driver) already passes. The inline comment block documenting why docker is required is a nice touch for future maintainers. Approve.

---

[github-actions]

Claude finished @alexenica's task in 2m 17s —— View job

---

Coder-eval task lint (advisory)

2 task YAMLs changed; verdict counts: 0 Critical, 1 High, 0 Medium, 0 Low, 2 OK.

Rubric: .claude/commands/lint-task.md. This check is advisory and never blocks merge.

Evidence of passing run

❌ High — PR body does not claim the changed tasks have been run and passed under the new driver: docker configuration. The body verifies the problem ("inspecting the 2026-07-15 run: of 778 tasks, only these two hit the error") but does not assert a passing run of the fixed tasks. Please edit the PR description to add a line like:

Ran skill-platform-queue-progress-e2e and skill-platform-trigger-api-e2e locally and they passed.

Per-task lint

tests/tasks/uipath-platform/resources/queue_progress_e2e.yaml — verdict: OK

No issues. Success criteria use run_command with an external Python check script (check_queue_progress.py) that queries tenant state directly — strong e2e validation. Prompt states the goal without prescribing procedure.

tests/tasks/uipath-platform/resources/trigger_api_e2e.yaml — verdict: OK

No issues. Same pattern as above — run_command with check_trigger_api.py querying tenant state. Prompt describes desired end-state, not procedure.

Within-PR duplicates

No duplicate clusters detected. The two tasks cover different resource types (queues vs. triggers) with different operations and check scripts.

Conclusion

⚠ 1 issue, max severity High (missing passing-run claim). Both tasks are well-structured with strong tenant-state validation. The driver: tempdir → driver: docker change is the only modification and the rationale comment is clear. Advisory only — not blocking merge.

---