Best possible solution:

Keep one canonical delivery-evidence classifier, retain both focused classifier and terminal-recovery coverage, and merge only after the responsible reviewer confirms that nested provider message.id is a successful-delivery receipt for the gateway result contract.

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

Yes. The linked issue gives a focused current-main reproduction using a2a-message-tool-mirror-dedupe, with one sessions_send, two successful message sends, and two requester-visible markers before this patch.

Is this the best way to solve the issue?

Yes, based on the supplied diff and QA evidence. Extending the existing delivery-evidence classifier is narrower and less error-prone than changing empty-response recovery or introducing a parallel gateway-specific path.

AGENTS.md: found and applied where relevant.

Codex review notes: model internal, reasoning high; reviewed against 0483001ef65c.

Label justifications:

• P1: The linked current-main reproduction shows a user-visible A2A reply can be sent twice after a successful first delivery.
• merge-risk: 🚨 message-delivery: The patch changes the success signal that determines whether recovery sends another visible message.
• rating: 🐚 platinum hermit: Overall readiness is 🐚 platinum hermit; proof is 🦞 diamond lobster and patch quality is 🐚 platinum hermit.
• status: 👀 ready for maintainer look: ClawSweeper has no concrete contributor-facing blocker left for this PR. Sufficient (linked_artifact): The PR body links an after-fix QA Gateway child and qa-channel scenario run that exercises the eight-second duplicate window and reports the improved result; no private credentials or identifiers are required in the supplied evidence.
• proof: sufficient: Contributor real behavior proof is sufficient. The PR body links an after-fix QA Gateway child and qa-channel scenario run that exercises the eight-second duplicate window and reports the improved result; no private credentials or identifiers are required in the supplied evidence.

PR surface:

Source +5, Tests +30. Total +35 across 3 files.

What I checked:

• Narrow delivery-classifier change: The proposed classifier adds message.id from a nested result object alongside existing top-level message, poll, and receipt delivery evidence. (src/agents/embedded-agent-message-tool-source-reply.ts:84, 5d47341e06f7)
• Classifier coverage: The focused classifier test asserts that a gateway-style result with details.message.id is considered delivered. (src/agents/embedded-agent-message-tool-source-reply.test.ts:319, 5d47341e06f7)
• Terminal recovery coverage: The terminal message-tool hook test covers the same nested gateway result, confirming it completes a message-tool-only source reply rather than invoking recovery. (src/agents/embedded-agent-runner/run/message-tool-terminal.test.ts:54, 5d47341e06f7)
• Observed problem and after-fix proof: The linked report records a current-main QA reproduction with one sessions_send followed by two successful message sends; the PR body links an after-fix QA Gateway/qa-channel run with the duplicate window scenario passing. (5d47341e06f7)
• Current-head provenance: The active branch head is the single commit fix(agents): recognize nested message delivery receipts; it is not part of the supplied current-main SHA, so this remains an open merge candidate rather than an implemented-on-main close. (5d47341e06f7)

Likely related people:

• steipete: The available PR and linked-bug context ties this contributor directly to the current A2A delivery reproduction, focused QA scenario, and proposed classifier repair; current-main feature ownership could not be established from the supplied read-only context. (role: recent area contributor; confidence: low; commits: 5d47341e06f7; files: src/agents/embedded-agent-message-tool-source-reply.ts, src/agents/embedded-agent-runner/run/message-tool-terminal.test.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.

• reviewed 2026-07-18T23:01:24.119Z sha 239488d :: needs maintainer review before merge. :: none
• reviewed 2026-07-18T23:18:10.593Z sha b981c31 :: needs maintainer review before merge. :: none