Best possible solution:

Keep current-main first-party Codex marketplace support and SMS runtime status projection unless maintainers approve a targeted rollback with migration/docs/proof for the user-visible fallout.

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

Yes. Source inspection of the PR head shows non-curated Codex marketplaces are filtered out and SMS runtime fields are omitted; I did not run live behavior because this was a read-only review.

Is this the best way to solve the issue?

No. A broad revert is not the best fix because current main matches upstream Codex marketplace names and preserves SMS health state; any real concern should be handled by a targeted rollback or policy change with upgrade proof.

Full review comments:

• [P1] Keep the first-party Codex marketplace allowlist — extensions/codex/src/app-server/config.ts:63
  This narrows native Codex plugin config back to openai-curated only. Current main accepts openai-bundled and openai-primary-runtime, and the upstream Codex checkout names those first-party marketplaces, so existing current-main configs for chrome, computer-use, documents, etc. would be dropped before inventory or activation can see them.
  Confidence: 0.95
• [P1] Preserve non-curated plugin app policy contexts — extensions/codex/src/app-server/session-binding.ts:260
  This parser now accepts only openai-curated policy context entries. Sessions saved on current main with bundled or primary-runtime plugin apps will read back without pluginAppPolicyContext, losing the per-plugin destructive-action and MCP server policy state used by later Codex thread/app handling.
  Confidence: 0.91
• [P2] Preserve SMS runtime fields in status snapshots — extensions/sms/src/channel.ts:259-267
  The replacement SMS snapshot hook ignores the runtime argument and returns no running, connected, last error, or activity timestamps. Gateway status passes the live runtime snapshot into this hook, and health evaluation treats missing running as not-running, so a configured running SMS account can appear unhealthy or stopped.
  Confidence: 0.92

Overall correctness: patch is incorrect
Overall confidence: 0.94

AGENTS.md: found and applied where relevant.

Codex review notes: model gpt-5.5, reasoning high; reviewed against 7b82901e58d3.

Label justifications:

• P1: The PR would break current-main Codex plugin configuration/session behavior and SMS health reporting for real gateway users if merged.
• merge-risk: 🚨 compatibility: The diff narrows a current-main user-facing Codex config enum and docs back to openai-curated-only, invalidating existing non-curated entries.
• merge-risk: 🚨 session-state: The diff changes session-binding parsing so saved plugin app policy contexts for bundled or primary-runtime marketplaces are discarded.
• merge-risk: 🚨 availability: The diff removes SMS runtime fields used by health/readiness, causing active configured SMS accounts to appear not-running.
• rating: 🧂 unranked krab: Overall readiness is 🧂 unranked krab; proof is 🧂 unranked krab and patch quality is 🧂 unranked krab.
• status: 📣 needs proof: The PR needs real behavior proof before ClawSweeper can clear the contributor ask. Needs real behavior proof before merge: The PR body lists static checks and local source inspection only; it needs redacted terminal/live output from a real after-revert Codex or SMS setup, then the contributor should update the PR body to trigger re-review or ask a maintainer to comment @clawsweeper re-review.

PR surface:

Source -99, Tests -204, Docs -67. Total -370 across 22 files.

What I checked:

• Repository policy applied: Root and scoped repository policy were read fully; the Codex dependency gate and compatibility/session-state review rules directly affected this verdict. (AGENTS.md:1, 7b82901e58d3)
• Current main accepts all first-party Codex marketplaces: Current main defines CODEX_PLUGINS_MARKETPLACE_NAMES with openai-curated, openai-bundled, and openai-primary-runtime, and resolveCodexPluginsPolicy keeps entries whose marketplaceName is in that allowlist. (extensions/codex/src/app-server/config.ts:70, 7b82901e58d3)
• PR head narrows Codex config back to one marketplace: The PR head replaces the allowlist with a single CODEX_PLUGINS_MARKETPLACE_NAME constant and drops non-curated pluginPolicies before inventory or activation can see them. (extensions/codex/src/app-server/config.ts:63, 3595795e7e14)
• Codex upstream names bundled and primary-runtime marketplaces: The required sibling Codex checkout shows openai-curated and openai-bundled constants, the CLI names openai-primary-runtime, and implicit system-root handling recognizes that primary-runtime marketplace path. (../codex/codex-rs/cli/src/plugin_cmd.rs:29, 0eb7e6d79b44)
• PR head rejects saved non-curated plugin policy contexts: The PR head session-binding parser accepts only openai-curated in pluginAppPolicyContext entries, so bundled or primary-runtime policy context saved by current main reads back as undefined. (extensions/codex/src/app-server/session-binding.ts:260, 3595795e7e14)
• Current main preserves SMS runtime fields for health: Current main's SMS snapshot helper carries running, connected, lastConnectedAt, lastError, lastInboundAt, and lastOutboundAt, while gateway health treats missing or false running as not-running. (extensions/sms/src/channel.ts:44, 7b82901e58d3)

Likely related people:

• Vincent Koc: Current-main blame for the Codex allowlist and SMS status snapshot helper points to commit d31f4e2, which recently touched the central files now being reverted. (role: recent area contributor; confidence: medium; commits: d31f4e2d6235; files: extensions/codex/src/app-server/config.ts, extensions/sms/src/channel.ts)
• yaanfpv: The linked merged PR that introduced the first-party Codex marketplace support is credited to this GitHub author in the provided GitHub context; the merge commit is the behavior this PR attempts to revert. (role: related fix PR author; confidence: medium; commits: 4f3d8a57ddd3; files: extensions/codex/src/app-server/config.ts, extensions/codex/src/app-server/plugin-inventory.ts, extensions/codex/src/app-server/session-binding.ts)

• 🦀 challenger crab: rare, exceptional readiness with strong proof, clean implementation, and convincing validation.
• 🦞 diamond lobster: very strong readiness with only minor maintainer review expected.
• 🐚 platinum hermit: good normal PR, likely mergeable with ordinary maintainer review.
• 🦐 gold shrimp: useful signal, but proof or patch confidence is still limited.
• 🦪 silver shellfish: thin signal; proof, validation, or implementation needs work.
• 🧂 unranked krab: not merge-ready because proof is missing/unusable or there are serious correctness or safety concerns.
• 🌊 off-meta tidepool: rating does not apply to this item.

Shiny media proof means a screenshot, video, or linked artifact directly shows the changed behavior. Runtime, network, CSP, and security claims still need visible diagnostics.

• ClawSweeper keeps one durable marker-backed review comment per issue or PR.
• Re-runs edit this comment so the latest verdict, findings, and automation markers stay together instead of adding duplicate bot comments.
• A fresh review can be triggered by eligible @clawsweeper re-review comments, exact-item GitHub events, scheduled/background review runs, or manual workflow dispatch.
• PR/issue authors and users with repository write access can comment @clawsweeper re-review or @clawsweeper re-run on an open PR or issue to request a fresh review only.
• Maintainers can also comment @clawsweeper review to request a fresh review only.
• Fresh-review commands do not start repair, autofix, rebase, CI repair, or automerge.
• Maintainer-only repair and merge flows require explicit commands such as @clawsweeper autofix, @clawsweeper automerge, @clawsweeper fix ci, or @clawsweeper address review.
• Maintainers can comment @clawsweeper explain to ask for more context, or @clawsweeper stop to stop active automation.