Best possible solution:

Land the narrow helper-based fix after the contributor adds real behavior proof showing a long startup memory or heartbeat exec-output truncation with emoji stays well-formed, or after a maintainer explicitly overrides the proof gate.

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

Yes, source inspection gives a high-confidence current-main reproduction path: call the affected truncation functions with content whose cutoff lands between a surrogate pair, and raw slice() can return a dangling surrogate. I did not run tests locally because this review was read-only.

Is this the best way to solve the issue?

Yes, this is the narrowest maintainable fix shape: use the existing truncateUtf16Safe helper at the two raw prefix-truncation sites while preserving current limits and markers. The PR tests are useful, though real behavior proof is still missing.

AGENTS.md: found and applied where relevant.

Codex review notes: model internal, reasoning high; reviewed against 4fa3118049c1.

Label changes:

• add P2: This is a focused bug fix for malformed model/user-visible text at truncation boundaries with limited blast radius and no evidence of an urgent live outage.
• add rating: 🦪 silver shellfish: Overall readiness is 🦪 silver shellfish; proof is 🦪 silver shellfish and patch quality is 🐚 platinum hermit.
• add 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 and CI provide unit-test/check evidence only; before merge, the contributor should add a redacted live run, terminal output, logs, or equivalent runtime proof showing the changed truncation behavior after the fix. 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.

Label justifications:

• P2: This is a focused bug fix for malformed model/user-visible text at truncation boundaries with limited blast radius and no evidence of an urgent live outage.
• rating: 🦪 silver shellfish: Overall readiness is 🦪 silver shellfish; proof is 🦪 silver shellfish and patch quality is 🐚 platinum hermit.
• 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 and CI provide unit-test/check evidence only; before merge, the contributor should add a redacted live run, terminal output, logs, or equivalent runtime proof showing the changed truncation behavior after the fix. 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 +2, Tests +38. Total +40 across 4 files.

What I checked:

• Current startup-context behavior: trimStartupMemoryContent trims startup memory and truncates over-limit content with raw String.prototype.slice, so a cutoff inside a surrogate pair can produce a dangling surrogate in the model-visible startup prelude. (src/auto-reply/reply/startup-context.ts:92, 4fa3118049c1)
• Current heartbeat exec-event behavior: buildExecEventPrompt still truncates long exec event text with raw rawEventText.slice(0, MAX_EXEC_EVENT_PROMPT_CHARS) before appending the truncation marker. (src/infra/heartbeat-events-filter.ts:124, 4fa3118049c1)
• Existing helper contract: truncateUtf16Safe delegates to sliceUtf16Safe, which adjusts the end boundary when the limit would leave a dangling high surrogate; the helper tests include truncating at a surrogate-pair boundary. (packages/normalization-core/src/utf16-slice.ts:43, 4fa3118049c1)
• Startup caller path: Bare /new and /reset paths call buildSessionStartupContextPrelude, and buildReplyPromptEnvelopeBase places that prelude into the effective model body for a reset/startup turn. (src/auto-reply/reply/get-reply-run.ts:768, 4fa3118049c1)
• Heartbeat caller path: Heartbeat preflight collects exec completion events and passes them into buildExecEventPrompt, which becomes the heartbeat run prompt. (src/infra/heartbeat-runner.ts:1262, 4fa3118049c1)
• PR diff review: The live PR diff changes only the two truncation call sites plus adjacent tests; it adds no dependency, workflow, config, lockfile, storage, auth, or security-boundary surface. (53892cc55d97)

Likely related people:

• Takhoffman: GitHub path history shows Takhoffman introduced and refined startup-context runtime behavior, including startup memory loading, bounds, and UTC/local date handling. (role: startup context feature history; confidence: high; commits: 94340b959830, 4f00b769251d, 06ff594a3e99; files: src/auto-reply/reply/startup-context.ts)
• steipete: GitHub path history shows steipete split and maintained heartbeat event filtering and later extracted the normalization-core package that owns the UTF-16 helper used here. (role: heartbeat and normalization area contributor; confidence: high; commits: 5a431f57fc18, e2362d352d14, 65c9eddae87e; files: src/infra/heartbeat-events-filter.ts, packages/normalization-core/src/utf16-slice.ts)
• GodsBoy: GitHub path history ties GodsBoy to the exec completion payload behavior that feeds the heartbeat prompt path touched by this PR. (role: exec completion prompt contributor; confidence: medium; commits: 349749f73dee; files: src/infra/heartbeat-events-filter.ts)
• mmaps: GitHub path history shows mmaps extended exec completion detection for local background exec formats, which is adjacent to the prompt path under review. (role: exec completion detection contributor; confidence: medium; commits: 19a2e9ddb5a8; files: src/infra/heartbeat-events-filter.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.