-
-
Notifications
You must be signed in to change notification settings - Fork 80.7k
[Bug]: Track Codex harness stability work #66251
Copy link
Copy link
Closed as not planned
Labels
P2Normal backlog priority with limited blast radius.Normal backlog priority with limited blast radius.clawsweeper:fix-shape-clearClawSweeper found a clear likely implementation shape for this issue.ClawSweeper found a clear likely implementation shape for this issue.clawsweeper:needs-maintainer-reviewClawSweeper marked this issue as needing maintainer review before automation.ClawSweeper marked this issue as needing maintainer review before automation.clawsweeper:needs-product-decisionClawSweeper marked this issue as needing a product or behavior decision.ClawSweeper marked this issue as needing a product or behavior decision.clawsweeper:no-new-fix-prClawSweeper does not recommend queueing a new automated fix PR for this issue.ClawSweeper does not recommend queueing a new automated fix PR for this issue.impact:auth-providerAuth, provider routing, model choice, or SecretRef resolution may break.Auth, provider routing, model choice, or SecretRef resolution may break.impact:session-stateSession, memory, transcript, context, or agent state can drift or corrupt.Session, memory, transcript, context, or agent state can drift or corrupt.issue-rating: 🌊 off-meta tidepoolIssue quality rating does not apply to this item.Issue quality rating does not apply to this item.staleMarked as stale due to inactivityMarked as stale due to inactivity
Description
Metadata
Metadata
Assignees
Labels
P2Normal backlog priority with limited blast radius.Normal backlog priority with limited blast radius.clawsweeper:fix-shape-clearClawSweeper found a clear likely implementation shape for this issue.ClawSweeper found a clear likely implementation shape for this issue.clawsweeper:needs-maintainer-reviewClawSweeper marked this issue as needing maintainer review before automation.ClawSweeper marked this issue as needing maintainer review before automation.clawsweeper:needs-product-decisionClawSweeper marked this issue as needing a product or behavior decision.ClawSweeper marked this issue as needing a product or behavior decision.clawsweeper:no-new-fix-prClawSweeper does not recommend queueing a new automated fix PR for this issue.ClawSweeper does not recommend queueing a new automated fix PR for this issue.impact:auth-providerAuth, provider routing, model choice, or SecretRef resolution may break.Auth, provider routing, model choice, or SecretRef resolution may break.impact:session-stateSession, memory, transcript, context, or agent state can drift or corrupt.Session, memory, transcript, context, or agent state can drift or corrupt.issue-rating: 🌊 off-meta tidepoolIssue quality rating does not apply to this item.Issue quality rating does not apply to this item.staleMarked as stale due to inactivityMarked as stale due to inactivity
Type
Fields
Priority
None yet
Bug type
Behavior bug (incorrect output/state without crash)
Beta release blocker
No
Summary
This is a tracking issue for coordinating Codex harness stability work. Detailed repro steps, root-cause notes, and implementation discussion should live in linked child issues or PRs.
Steps to reproduce
This is an umbrella tracking issue rather than a single repro case.
The bundled Codex harness has several related reliability issues across gateway startup, Codex app-server initialization, session reset, native compaction, and token accounting. This issue is meant to organize Codex-specific reports and keep the stabilization work easy to follow.
Expected behavior
The bundled Codex harness should be stable enough to use as the primary runtime for Codex-backed OpenClaw agents.
At a high level:
Actual behavior
Multiple related Codex harness bugs have been observed or reported. This parent issue is only for coordination, ordering, cross-linking, and broader discussion about getting the Codex harness into a stable state.
OpenClaw version
2026.4.10 / 2026.4.11 reports, plus current
maininvestigationOperating system
macOS and Linux reports
Install method
npm/global CLI and local source investigation
Model
codex/gpt-5.4
Provider / routing chain
OpenClaw -> bundled codex plugin -> Codex app-server
Additional provider/model setup details
NOT_ENOUGH_INFO
Logs, screenshots, and evidence
Tracking map
1. Gateway startup, plugin activation, and model catalog
These cover cases where Codex appears configured or loaded, but the live gateway/model registry does not have the expected Codex runtime surface available.
codex/*provider catalog not registered during gateway startup —Unknown model: codex/gpt-5.4on every restart #64938 — Nativecodex/*provider catalog is not registered during gateway startup; users seeUnknown model: codex/gpt-5.4on restart.serviceTiercontract drift plus startup warmup using static model resolution when async resolution is needed.2. Codex app-server initialization and protocol robustness
These cover handshake, initialization, user-agent parsing, and app-server protocol fragility.
Codex Desktop/0.118.0.NODE_OPTIONS --requirepreloads.imagetool remains exposed. Duplicate: Codex-native vision turns can stall when inbound images are present and the dynamic image tool remains exposed #65052. Proposed fix: fix(codex): avoid image tool loops on vision turns and respect image-model providers #65061.3. Codex reset and native app-server resume state
These cover cases where Codex-backed reset/new-session behavior does not fully isolate the next run from old native Codex app-server state.
4. Native Codex compaction, token accounting, and context reporting
These cover misleading context size and Codex-specific compaction/status accounting.
Scope notes
Please keep this tracker focused on the bundled Codex harness and Codex app-server integration. Generic session, compaction, transcript replay, CLI fallback, or ACP issues should only be linked here when they have a clear Codex-harness-specific failure mode.
Notes for contributors
Please prefer focused child issues and PRs over adding detailed repros directly here. A good child issue should include:
Impact and severity
Affected: Users running OpenClaw with the bundled Codex harness.
Severity: High for Codex-backed long-running sessions.
Frequency: Multiple related issues have been observed across startup, reset, status, and compaction flows.
Consequence: Codex-backed sessions can fail to start through the expected harness, preserve stale native context after reset, or report misleading context usage after compaction.
Additional information
This issue is intended as a coordination hub. Please keep detailed reproduction steps, root-cause notes, and implementation discussion in the linked child issues or PRs where possible.