Best possible solution:

Keep the focused prompt-response usage persistence and regression tests, with release notes handled by the release-owned process.

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

Yes, source-level reproduction is high confidence: on current main, usage normalization only reads usage_update payloads and runPromptTurn drops successful prompt response usage before session persistence. I did not run a live adapter session in this read-only review.

Is this the best way to solve the issue?

Yes, recording prompt-response usage after the update drain reuses the existing token normalization and persisted/status surfaces without adding new config or public naming conventions.

AGENTS.md: found and applied where relevant.

Codex review notes: model internal, reasoning high; reviewed against a6907746b5bd.

Label changes:

• add rating: 🐚 platinum hermit: Overall readiness is 🐚 platinum hermit; proof is 🌊 off-meta tidepool and patch quality is 🐚 platinum hermit.
• add status: 👀 ready for maintainer look: ClawSweeper has no concrete contributor-facing blocker left for this PR. Not applicable: This is a ClawSweeper-generated bot autofix PR, so the external-contributor real behavior proof gate does not apply; regression tests and CI still validate the behavior.
• remove status: ⏳ waiting on author: Current PR status label is status: 👀 ready for maintainer look.
• remove rating: 🦐 gold shrimp: Current PR rating is rating: 🐚 platinum hermit, so this older rating label is no longer current.
• remove merge-risk: 🚨 other: Current PR review selected no merge-risk labels.

Label justifications:

• P2: The PR fixes a normal-priority persisted session-state bug with bounded runtime and test changes.
• rating: 🐚 platinum hermit: Overall readiness is 🐚 platinum hermit; proof is 🌊 off-meta tidepool and patch quality is 🐚 platinum hermit.
• status: 👀 ready for maintainer look: ClawSweeper has no concrete contributor-facing blocker left for this PR. Not applicable: This is a ClawSweeper-generated bot autofix PR, so the external-contributor real behavior proof gate does not apply; regression tests and CI still validate the behavior.

What I checked:

• Target policy read: AGENTS.md was read fully; its bug-fix evidence, runtime validation, and changelog/release-note guidance apply to this runtime/session PR. (AGENTS.md:220, a6907746b5bd)
• Current-main source reproduction: On current main, token usage normalization reads from usage_update._meta.usage or top-level usage-update fields only, so { used, size } updates produce no persisted token breakdown. (src/session/conversation-model.ts:445, a6907746b5bd)
• Current-main prompt usage drop: On current main, runPromptTurn awaits the prompt response but returns only stopReason/source, so successful PromptResponse.usage is not folded into session persistence. (src/runtime/engine/prompt-turn.ts:24, a6907746b5bd)
• Existing status dependency: getStatus().usage is built from persisted cumulative_token_usage and request_token_usage, matching the linked issue's empty-status symptom when those fields are never populated. (src/runtime/engine/manager.ts:329, a6907746b5bd)
• Adjacent response usage path: The CLI output formatter already reads final JSON-RPC result.usage, confirming the response-side source is part of current behavior. (src/cli/output/output.ts:1190, a6907746b5bd)
• PR implementation path: The PR records response.usage after the post-success update drain, so prompt-response usage reaches the session conversation before persistence. (src/runtime/engine/prompt-turn.ts:41, 62a39174d78e)

Likely related people:

• DaniAkash: Authored the merged runtime/status usage work that added persisted usage exposure through the runtime manager, public contract, reducer, and tests. (role: feature introducer; confidence: high; commits: f6de6dd18aae; files: src/runtime/engine/manager.ts, src/runtime/public/contract.ts, src/session/conversation-model.ts)
• steipete: Authored the quiet-output prompt-result usage path and the release snapshot that introduced the current prompt-turn and conversation-model files on main. (role: adjacent runtime/output contributor; confidence: high; commits: ea2c1cfb5ab3, 9149b26f4971; files: src/cli/output/output.ts, src/runtime/engine/prompt-turn.ts, src/session/conversation-model.ts)
• Onur Solmaz: Introduced the runtime embedding API and central conversation/runtime manager surfaces that this persistence/status path builds on. (role: runtime API introducer; confidence: medium; commits: be510ba918d4; files: src/runtime/engine/manager.ts, src/runtime/public/contract.ts, src/session/conversation-model.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.