Best possible solution:

Land the shared UTF-16 helper implementation after refreshing it against current main, retaining the existing cap, omission marker, redaction order, head/tail layout, and focused boundary tests.

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

Yes—the raw prefix/head/tail slicing can deterministically split an emoji by positioning its surrogate pair across the configured cap boundary; the supplied production-path run verifies the repaired output, although an unpatched current-main live run was not executed in this review.

Is this the best way to solve the issue?

Yes—the PR applies the repository’s existing UTF-16-safe normalization helpers at the actual downstream cap instead of adding a parallel truncation implementation, while preserving the established budget and layout.

AGENTS.md: found and applied where relevant.

Codex review notes: model internal, reasoning high; reviewed against 154d53c4f6fc.

Label changes:

• add proof: sufficient: Contributor real behavior proof is sufficient. The linked exact-head workflow demonstrates the repaired production SQLite-to-provider path, including the deliberate surrogate boundary, bounded transcript, omission marker, two live loopback provider requests, and a successful skill_workshop completion round trip without a dangling surrogate.
• add rating: 🦞 diamond lobster: Overall readiness is 🦞 diamond lobster; proof is 🦞 diamond lobster and patch quality is 🦞 diamond lobster.
• add status: 👀 ready for maintainer look: ClawSweeper has no concrete contributor-facing blocker left for this PR. Sufficient (linked_artifact): The linked exact-head workflow demonstrates the repaired production SQLite-to-provider path, including the deliberate surrogate boundary, bounded transcript, omission marker, two live loopback provider requests, and a successful skill_workshop completion round trip without a dangling surrogate.
• remove rating: 🦪 silver shellfish: Current PR rating is rating: 🦞 diamond lobster, so this older rating label is no longer current.
• remove status: 📣 needs proof: Current PR status label is status: 👀 ready for maintainer look.

Label justifications:

• P2: Malformed provider-bound transcript text is a real bounded Skill Workshop bug with focused scope and no evidence of broad runtime outage or data loss.
• rating: 🦞 diamond lobster: Overall readiness is 🦞 diamond lobster; proof is 🦞 diamond lobster and patch quality is 🦞 diamond lobster.
• status: 👀 ready for maintainer look: ClawSweeper has no concrete contributor-facing blocker left for this PR. Sufficient (linked_artifact): The linked exact-head workflow demonstrates the repaired production SQLite-to-provider path, including the deliberate surrogate boundary, bounded transcript, omission marker, two live loopback provider requests, and a successful skill_workshop completion round trip without a dangling surrogate.
• proof: sufficient: Contributor real behavior proof is sufficient. The linked exact-head workflow demonstrates the repaired production SQLite-to-provider path, including the deliberate surrogate boundary, bounded transcript, omission marker, two live loopback provider requests, and a successful skill_workshop completion round trip without a dangling surrogate.

PR surface:

Source +1, Tests +31. Total +32 across 2 files.

What I checked:

• Affected source path: The unpatched per-session cap uses raw String.slice operations for its prefix, retained head, and retained tail, allowing a cap boundary to split a UTF-16 surrogate pair. (src/skills/workshop/history-scan-transcript-content.ts:23, 86772735ac14)
• Fix shape: The PR replaces each boundary-sensitive slice with the existing normalization-core helpers and computes the tail budget from the actual safe head length, preserving the overall cap and omission layout. (src/skills/workshop/history-scan-transcript-content.ts:1, 917f5f2d001e)
• Regression coverage: Three tests exercise the prefix-only, retained-head, and retained-tail boundaries, assert the original character budget and omission behavior, and reject dangling surrogates. (src/skills/workshop/history-scan.test.ts:248, 1ba1110785b7)
• Real behavior proof: The linked exact-head workflow persisted a real SQLite session, followed the production candidate/read/review path, made two loopback provider requests, retained the omission marker, and reported danglingSurrogate=false with a completed skill_workshop round trip. (src/skills/workshop/history-scan-transcript-content.ts:17, 1ba1110785b7)
• Related work is distinct: The open related PR addresses earlier CLI and 60,000-character experience-review truncation surfaces; this PR addresses the later 16,000-character per-session history-scan cap, so neither safely supersedes the other. (src/skills/workshop/history-scan-transcript-content.ts:17, 1ba1110785b7)
• Validation state: The exact-head CI gate, focused regression tests, formatting check, production types, lint, build artifacts, security-sensitive guard, and real behavior proof completed successfully. (1ba1110785b7)

Likely related people:

• vincentkoc: Recent merged history connects this contributor to the Skill Workshop surface, making them a plausible reviewer for workshop behavior, although the exact history-scan ownership trail is ambiguous. (role: adjacent Skill Workshop contributor; confidence: low; commits: 077d813b0; files: src/skills/workshop/)

• 🦀 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 2026-07-15T07:35:33.558Z sha 1ba1110 :: needs real behavior proof before merge. :: none
• reviewed 2026-07-15T08:51:27.422Z sha 1ba1110 :: needs real behavior proof before merge. :: none
• reviewed 2026-07-15T10:05:02.259Z sha 1ba1110 :: needs real behavior proof before merge. :: none