Best possible solution:

Port any still-useful guard to current embedded-agent-runner paths, reconcile it with the active completion delivery contract, and require redacted real runtime or transport proof before merge.

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

Yes at source level, but not with live runtime proof yet: current main lacks the proposed sessions_yield abort acceptance gate, and the active completion path can still queue into an active requester before falling back.

Is this the best way to solve the issue?

No as submitted: the guard is on obsolete files and the direct-handoff semantics need maintainer agreement against the current delivery contract.

Full review comments:

• [P1] Port the queue guard to embedded-agent-runner — src/agents/pi-embedded-runner/run/attempt.ts:3748
  Current main no longer has src/agents/pi-embedded-runner, so adding isAcceptingMessages only to that old handle cannot protect the current queue path. Please port the type, handle state, registry guard, and regression coverage to src/agents/embedded-agent-runner.
  Confidence: 0.95

Overall correctness: patch is incorrect
Overall confidence: 0.9

AGENTS.md: found and applied where relevant.

Codex review notes: model internal, reasoning high; reviewed against db2488b6e3d6.

Label justifications:

• P1: The PR targets a yielded parent/subagent completion workflow that can freeze or misdeliver real agent-session work.
• merge-risk: 🚨 session-state: The diff changes whether an active yielded requester session accepts queued completion input during the abort transition.
• merge-risk: 🚨 message-delivery: The diff changes subagent completion delivery ordering and fallback behavior, which could drop, duplicate, or misroute completion output.
• 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 external PR body lists tests only and does not include redacted after-fix runtime logs, terminal output, copied live output, screenshots, recordings, or linked artifacts showing the sessions_yield completion path. After adding proof, update the PR body; ClawSweeper should re-review automatically. If it does not, the PR author or someone with repository write access can comment @clawsweeper re-review.

PR surface:

Source +37, Tests +8. Total +45 across 6 files.

What I checked:

• Repository policy applied: Root AGENTS.md and scoped agent-runner AGENTS.md files were read fully and applied to the deep current-main, sibling-surface, proof, and merge-risk review. (AGENTS.md:1, db2488b6e3d6)
• PR targets removed runner path: The submitted diff modifies four files under src/agents/pi-embedded-runner, while current main no longer has that tree. (f9a83962ad7e)
• Current queue handle lacks acceptance probe: Current EmbeddedAgentQueueHandle has streaming, compacting, transcript-wait, cancel/abort, and source-reply mode fields, but no isAcceptingMessages equivalent for the sessions_yield abort transition. (src/agents/embedded-agent-runner/run-state.ts:19, db2488b6e3d6)
• Current queue eligibility lacks the PR guard: Current prepareEmbeddedAgentQueueMessage rejects missing, non-streaming, compacting, transcript-wait-unsupported, and source-reply mismatched handles, but not a streaming run that is in the sessions_yield abort window. (src/agents/embedded-agent-runner/runs.ts:365, db2488b6e3d6)
• Current yield callback has no abort-in-progress state: The current onYield callback records yield, queues the interrupt, aborts with sessions_yield, and aborts the session, but does not flip a queue-acceptance state like the PR proposes. (src/agents/embedded-agent-runner/run/attempt.ts:1347, db2488b6e3d6)
• Active completion wake still queues active requesters: Current completion delivery still attempts resolveActiveWakeWithRetries for active completion requesters and reports path: "steered" when the queue outcome is accepted. (src/agents/subagent-announce-delivery.ts:1398, db2488b6e3d6)

Likely related people:

• vincentkoc: Recent GitHub path history shows multiple current subagent completion handoff and embedded runner queue changes in the affected files. (role: recent area contributor; confidence: high; commits: 81eaa88ce56d, 68a1e00b73bd, 7fc4bbc0bcba; files: src/agents/embedded-agent-runner/runs.ts, src/agents/subagent-announce-delivery.ts)
• steipete: Path history shows the split of subagent announce delivery helpers and prior work shaping the current delivery owner boundary. (role: feature-history contributor; confidence: medium; commits: b75be0914491; files: src/agents/subagent-announce-delivery.ts, src/agents/subagent-announce-dispatch.ts)
• jriff: The cooperative sessions_yield tool involved in this wake/freeze path traces to the sessions_yield feature introduction. (role: introduced behavior; confidence: high; commits: 3fa91cd69d5d; files: src/agents/tools/sessions-yield-tool.ts)
• joshavant: Prior commits changed subagent completion announce delivery timing and subagent DM completion delivery in the same delivery module. (role: prior delivery timing contributor; confidence: medium; commits: 903d9c13f3ee, 0896917e1ad9; files: src/agents/subagent-announce-delivery.ts, src/agents/subagent-announce-delivery.test.ts)
• IWhatsskill: Recent embedded runner history includes transcript flush behavior around after-turn cleanup, adjacent to the sessions_yield abort/flush window. (role: recent adjacent contributor; confidence: medium; commits: 7ac8b48a08bf; files: src/agents/embedded-agent-runner/run/attempt.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.

• The egg game starts only after the PR passes the real-behavior proof check.
• Before that, no creature or rarity is rolled. The treat waits for real proof.
• This is still just collectible flavor: proof affects review readiness, not creature quality.