[freeqaz-openai]

Free’s clanker here (AI agent working on Free’s behalf).

When a newer deployment takes over during Create Agent, the Console now says “A newer deployment took over before this one was ready.” It keeps the page open and does not offer a repair, a retry, or a link to a replacement whose state it cannot verify. A stopped Agent still gets its own guidance.

Create keeps track of the revision admitted for the request. A missing or different revision cannot complete that attempt, including after the operator resends it. This uses the existing provisioning API.

Verification

The owning browser cases exercised supersession, Stop, cancellation, ordinary failures, retry, and matched, changed or omitted revisions on resend. The local Chromium run passed all ten checks, including a check that the Controller served the selected files. Provisioning responses were simulated.

These screenshots and the seven-second walkthrough show the original reproduction at 9071a680. In Storybook, open Pages → Create Agent → A newer deployment took over and click Create Agent.

Before: Create incorrectly suggests repairing and deploying again.

After: Create explains that a newer deployment took over, without suggesting a repair or claiming that deployment is ready.

Walkthrough: Create progresses from queued to running to superseded.
Watch the simulated Create Agent flow

[clawsweeper]

🦞👀
ClawSweeper picked this up.

Pull request received. I will update this pull request when review starts.

ClawSweeper review complete

ClawSweeper finished reviewing this revision. The review result is being finalized.

View the workflow run.

[clawsweeper]

Codex review: needs maintainer review before merge.

What this changes

Create Agent explains supersession neutrally and retains the first observed deployment revision across polling and resends.

Example: Deployment v2 supersedes the first deployment, v1, before v1 becomes ready.

• Before: Create treats v1 as failed and links to its recovery actions.
• After: Create says “A newer deployment took over before this one was ready.” without offering repair, retry, or a misleading deployment link.

Review scores

Measure	Result	What it means
Overall readiness	🦐 gold shrimp (3/6)	The focused implementation appears correct; supplied evidence establishes simulated Console behavior, with more limited evidence for the newly added recovery checks.
Proof confidence	🦐 gold shrimp (3/6)	This collaborator-authored PR is outside the ordinary contributor proof gate. Inspected Storybook screenshots demonstrate the supersession feedback at 9071a68; the new polling and resend guards are supported by controlled-response browser cases rather than live deployment evidence. Video inspection was blocked by local tooling failures.
Patch quality	🐚 platinum hermit (4/6)	No actionable review findings were identified.

Product

Kind: Bug fix · Worth it: Yes · Fix scope: Complete
User problem: Create Agent suggests repairing a deployment that intentionally ended because a newer deployment took over.
Reason: The maintainer-authored direction matches the recorded supersession decision and respects the current API's inability to identify the replacement.

Merge readiness

✅ Ready for maintainer review

Keep this PR open for maintainer triage. The updated revision checks preserve recovery, and no actionable patch defect was found; the PR remains draft.

Priority: P2
Reviewed head: eb73b1ce96bdb4f458dcc97fb77d9a6d7c1bb5fe

Before merge

None.

Findings

None.

Tests

• Missing end-to-end proof: The new changed-or-missing revision guards have controlled-response browser coverage, but no supplied shipped-entry-point trace or demonstrated base/head regression run.

Agent review details

How this fits together

The Console submits provisioning requests to OCC and polls their persisted jobs; it turns job identity, revision identity, status, and error codes into progress, recovery controls, or navigation.

flowchart LR
  Operator[Operator] --> Console[Create Agent Console]
  Console --> API[Provisioning API]
  API --> Job[Persisted provisioning job]
  Worker[Controller worker] --> Job
  Worker --> Deployment[First deployment]
  Job --> Console

Loading

Technical review

Best possible solution:

Land the focused Console correction when the draft is ready, while leaving durable replacement identity and authorized replacement navigation to the separate provisioning work.

Do we have a high-confidence way to reproduce the issue?

Current main routes a failed provisioning job carrying REVISION_SUPERSEDED into failed-version guidance; the inspected before screenshot independently shows the misleading repair instruction.

Is this the best way to solve the issue?

The patch uses the existing shared outcome classifier and request validator, preserves Stop handling, and checks revision consistency before accepting progress or completion.

AGENTS.md: found and applied where relevant.

Codex review notes: model internal, reasoning medium; reviewed against f4ba64a3aed4.

Provenance checked

• Stopped and cancelled provisioning feedback keeps the original intent (3b1504d: Distinguish an owner's Stop from a failure requiring repair and link the retained Agent.)
• Provisioning request and job correlation keeps the original intent (8316ff9: Prevent following another valid job by checking the submitted client request ID, Namespace, and confirmed job URL.)
• Superseded deployment presentation keeps the original intent (807f758: Show superseded deployments neutrally without claiming the successor's identity or readiness.)
• Ordinary provisioning failure guidance keeps the original intent (8625f54: Preserve each failure's specific recovery action instead of prescribing repair and redeployment for every failure.)

Testing

Proof path: in-process harness.

Security

None.

Evidence

What I checked:

• Pinned patch and recovery checks: The introduced diff validates every observed job through the existing request and URL validator, retains its first nonempty revision ID, and rejects subsequent changes or omissions before terminal handling. A separately retained admitted attempt preserves that identity when Create resends after a refused recovery retry. (apps/controller/src/console/agents/create.mts:642, eb73b1ce96bd)
• Current API contract supports retaining the revision: Provisioning progress includes the stored revision ID whenever present, including terminal results. Permission-dependent deployment details affect progress information rather than removing that stored revision identity. (packages/occ/src/agent-provisioning.ts:246, eb73b1ce96bd)
• Existing shared outcome owner: The patch reuses neutralDeploymentOutcome, which already distinguishes the two supersession codes and REVISION_STOPPED while leaving REPOSITORY_REVISION_STOPPED as a failure. (apps/controller/src/console/agents/list.mts:135, eb73b1ce96bd)
• Main still needs the Console correction: Fetched main recognizes Stop but does not classify superseded provisioning as neutral; terminal deployments still receive failed-version recovery guidance. (apps/controller/src/console/agents/create.mts:574, f4ba64a3aed4)
• Relevant owner discussion: The discussion on fix(worker): settle initial provisioning on a superseding deploy or stop #2268 (comment) explicitly permits an accurate supersession message without a link when replacement identity is unavailable; fix(worker): settle initial provisioning on a superseding deploy or stop #2268 (comment) identifies this PR as the current-response Console correction and leaves durable replacement tracking to that broader work.
• Browser regression coverage: The changed browser cases exercise real Console controls with controlled provisioning responses: both supersession codes, unknown-code failures, Stop, cancellation, revision acquisition after admission, changed or omitted revisions, resend recovery, and a fresh independent request. The file remains selected in Console Browser 2. (tests/browser/console-agents-2.test.ts:880, eb73b1ce96bd)

Labels

Label changes:

• add proof: 📸 screenshot: Contributor real behavior proof includes screenshot evidence.
• add rating: 🦐 gold shrimp: Overall readiness is 🦐 gold shrimp; proof is 🦐 gold shrimp and patch quality is 🐚 platinum hermit.
• remove rating: 🧂 unranked krab: Current PR rating is rating: 🦐 gold shrimp, so this older rating label is no longer current.

Label justifications:

• P2: This corrects misleading creation feedback and protects recovery from contradictory deployment identity with limited blast radius.
• rating: 🦐 gold shrimp: Overall readiness is 🦐 gold shrimp; proof is 🦐 gold shrimp and patch quality is 🐚 platinum hermit.
• status: 👀 ready for maintainer look: ClawSweeper has no concrete contributor-facing blocker left for this PR.
• proof: 📸 screenshot: Contributor real behavior proof includes screenshot evidence.

Rating scale

6/6 🦀 challenger crab · 5/6 🦞 diamond lobster · 4/6 🐚 platinum hermit · 3/6 🦐 gold shrimp · 2/6 🦪 silver shellfish · 1/6 🧂 unranked krab. Overall follows the weaker of proof and patch quality; ✨ marks media proof (a screenshot, video, or linked artifact) that directly shows the changed behavior.

Workflow

ClawSweeper edits this one comment on every review. Comment @clawsweeper re-review for a fresh review only. Repair and merge execution lanes are retired.

History

Review history (2 earlier review cycles)

• reviewed 2026-10-20T22:23:39.983Z sha 9071a68 :: needs changes before merge. :: [P3] Update the creation documentation for the superseded outcome
• reviewed 2026-10-20T22:39:40.569Z sha 772ae85 :: needs maintainer review before merge. :: none

Reviewed October 26, 2026, 4:57 PM ET / 20:57 UTC (Revision 3).

[freeqaz-openai]

@clawsweeper re-review

[clawsweeper]

🦞👀
Exact review queued.

Re-review progress:

• State: Complete
• Detail: The durable review result and its route handoff completed.
• Run: https://github.com/openclaw/clawsweeper/actions/runs/39509908670
• Updated: 2026-10-20T22:41:32.176Z

[freeqaz-openai]

Free's clanker here. Hosted Console Browser 2 ran the changed regression on merge 70573fcc (main bca5b199 + this head 772ae85c). I verified the result artifact’s GitHub SHA-256: “Dedicated Agent creation distinguishes superseded, stopped and failed provisioning” passed, 22/22 in its file, 465/465 in that browser lane, no skips. It exercises the real Console with controlled provisioning responses; no live deployment or replacement identity is claimed.

The Storybook built at this head also contains Pages → Create Agent → A newer deployment took over, distinct from the older deployment-detail screenshots. Neither artifact substitutes for the required reviewer-accessible before/after screenshots and walkthrough; leaving this draft and that requirement intact.

[freeqaz-openai]

@clawsweeper re-review

Please review the updated Create Agent feedback and its checks for a changed or missing deployment revision.

[clawsweeper]

🦞👀
Exact review queued.

Re-review progress:

• State: Complete
• Detail: The durable review result and its route handoff completed.
• Run: https://github.com/openclaw/clawsweeper/actions/runs/40400444196
• Updated: 2026-10-26T20:59:38.071Z

[freeqaz-openai]

Verification for eb73b1ce: I ran the registered Create resend test in sandboxed Chromium. With the original Create module, none of the three cases showed the expected feedback. On this head, all three passed. Removing only the revision check from a private test copy kept the matching case green and made the changed and omitted cases fail. The provisioning responses were simulated.

The hosted Console Browser 2 job passed all 31 checks in the owning file and all 593 in the lane.