fix(msteams): truncate reflection prompt on a UTF-16 boundary by WeeLi-009 · Pull Request #96578 · openclaw/openclaw

WeeLi-009 · 2026-06-24T22:38:54Z

Summary: The Teams feedback reflection prompt was truncated with a raw UTF-16 slice, so an emoji at the boundary became a lone surrogate fed into the LLM prompt. This uses truncateUtf16Safe to truncate on a code-point boundary.

What Problem This Solves

The Microsoft Teams feedback reflection prompt truncated the thumbed-down response with raw UTF-16 slicing. When the response crossed the 500-code-unit cap at an emoji or other surrogate-pair boundary, the prompt could include a lone surrogate.

Trigger input used for proof:

"a".repeat(499) + "🦞" + "b".repeat(20)
Old truncation with .slice(0, 500) keeps only the high surrogate d83e, then appends ...
That malformed string can fail downstream UTF-8/URI encoding and can produce replacement-character behavior in logs or transport text

Fix

extensions/msteams/src/feedback-reflection-prompt.ts now reuses the shared truncateUtf16Safe helper from openclaw/plugin-sdk/text-utility-runtime when capping the thumbed-down response before appending the reflection prompt ellipsis.

That keeps the existing 500-character range check and prompt escaping behavior, but avoids cutting a UTF-16 surrogate pair in half. If the boundary emoji does not fully fit, the helper trims before the pair and the prompt remains well-formed.

Evidence

Command, run from a temporary worktree on fix/msteams-reflection-truncate-surrogate:

node --import tsx demo.mts

The demo imports the real fixed function:

import { buildReflectionPrompt } from "./extensions/msteams/src/feedback-reflection-prompt.ts";

Terminal output:

input.length = 521
emoji code point = 0x1f99e
BEFORE raw slice tail = "aaaa\ud83e..."
BEFORE raw slice tail code units = 0061 0061 0061 0061 d83e 002e 002e 002e
BEFORE lone-surrogate = true
BEFORE encodeURIComponent = URIError
AFTER prompt excerpt tail = "aaaaa..."
AFTER prompt excerpt tail code units = 0061 0061 0061 0061 0061 002e 002e 002e
AFTER lone-surrogate = false
AFTER contains boundary emoji = false

The fixed buildReflectionPrompt output is well-formed for this surrogate-boundary emoji input: the lone-surrogate regex reports false, and the boundary emoji is dropped rather than split.

Tests

Focused Vitest coverage exercises the prompt builder regression cases:

does not split UTF-16 surrogate pairs when truncating a thumbed-down response
keeps a boundary emoji when it fully fits before the truncation cap

Validation command:

node scripts/run-vitest.mjs extensions/msteams/src/feedback-reflection.test.ts

Result:

Test Files  1 passed (1)
Tests       20 passed (20)
[test] passed 1 Vitest shard in 78.96s

clawsweeper · 2026-06-24T22:39:52Z

Codex review: needs maintainer review before merge. Reviewed June 26, 2026, 12:15 AM ET / 04:15 UTC.

Summary
The PR replaces raw .slice(0, 500) truncation in the Microsoft Teams feedback reflection prompt with truncateUtf16Safe and adds surrogate-pair regression tests.

PR surface: Source +1, Tests +25. Total +26 across 2 files.

Reproducibility: yes. Current main and v2026.6.10 use raw UTF-16 slicing at the 500-code-unit cap, and the PR body supplies a boundary input that produces a lone surrogate before the fix and well-formed output after it; I did not run tests because this review is read-only.

Review metrics: none identified.

Merge readiness
Overall: 🐚 platinum hermit
Proof: 🦞 diamond lobster
Patch quality: 🐚 platinum hermit
Result: ready for maintainer review.

Overall follows the weaker of proof and patch quality, so missing proof can cap an otherwise strong patch.

Rank-up moves:

[P2] Rerun or resolve the failed live status checks before merge.

Risk before merge

[P1] Live PR status still includes failed Real behavior proof runs and checks-node-compact-small-whole-2, so merge should wait for rerun or maintainer resolution even though the patch review is clean.

Maintainer options:

Decide the mitigation before merge
Merge the narrow SDK-helper reuse after maintainer review and clean required checks; keep the sibling parent-context UTF-16 fix in fix(msteams): keep truncated parent context text well-formed #96569 as separate work.
Pause or close
Do not merge this PR until maintainers decide whether the risk is worth taking.

Next step before merge

[P2] No automated code repair is indicated; maintainers should review the correct narrow patch and rerun or resolve the failed status checks before merge.

Security
Cleared: The diff only changes string truncation through an existing SDK helper and adds tests; it does not touch dependencies, workflows, secrets, package resolution, or code execution surfaces.

Review details

Best possible solution:

Merge the narrow SDK-helper reuse after maintainer review and clean required checks; keep the sibling parent-context UTF-16 fix in #96569 as separate work.

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

Yes. Current main and v2026.6.10 use raw UTF-16 slicing at the 500-code-unit cap, and the PR body supplies a boundary input that produces a lone surrogate before the fix and well-formed output after it; I did not run tests because this review is read-only.

Is this the best way to solve the issue?

Yes. Reusing the existing public truncateUtf16Safe helper is the narrowest maintainable fix for this prompt path; local surrogate logic or broader Teams refactors would add unnecessary drift.

AGENTS.md: found and applied where relevant.

Codex review notes: model internal, reasoning high; reviewed against a6a4652c708a.

Label changes

Label justifications:

P2: This is a focused Microsoft Teams prompt correctness fix with limited blast radius and no outage, data-loss, or security evidence.
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 (terminal): The PR body includes terminal proof from a temporary worktree importing the real fixed function and showing the after-fix prompt has no lone surrogate.
proof: sufficient: Contributor real behavior proof is sufficient. The PR body includes terminal proof from a temporary worktree importing the real fixed function and showing the after-fix prompt has no lone surrogate.

Evidence reviewed

PR surface:

Source +1, Tests +25. Total +26 across 2 files.

View PR surface stats

Area	Files	Added	Removed	Net
Source	1	2	1	+1
Tests	1	25	0	+25
Docs	0	0	0	0
Config	0	0	0	0
Generated	0	0	0	0
Other	0	0	0	0
Total	2	27	1	+26

What I checked:

Current main still has the raw slice: buildReflectionPrompt still truncates thumbedDownResponse with slice(0, MAX_RESPONSE_CHARS) on current main, so the central bug is not already implemented there. (extensions/msteams/src/feedback-reflection-prompt.ts:22, a6a4652c708a)
Latest release still has the raw slice: v2026.6.10 resolves to commit aa69b12d0086b631b139c1435c9621a5783e3a40 and has the same raw slice(0, MAX_RESPONSE_CHARS) prompt truncation. (extensions/msteams/src/feedback-reflection-prompt.ts:22, aa69b12d0086)
PR head uses the existing helper at the implicated branch: The PR head imports truncateUtf16Safe and uses it only in the over-limit thumbedDownResponse branch before appending the existing ellipsis. (extensions/msteams/src/feedback-reflection-prompt.ts:23, d4da0d3afaa9)
Helper contract matches the fix: truncateUtf16Safe delegates to sliceUtf16Safe, whose end-boundary logic backs up when the cut would leave a high surrogate before a low surrogate. (src/utils.ts:100, a6a4652c708a)
Plugin SDK boundary is valid: text-utility-runtime exports truncateUtf16Safe, and package exports expose openclaw/plugin-sdk/text-utility-runtime, matching the extensions boundary policy. (src/plugin-sdk/text-utility-runtime.ts:21, a6a4652c708a)
Caller path inspected: runFeedbackReflection passes thumbedDownResponse and userComment into buildReflectionPrompt, so the changed helper sits on the actual background reflection prompt path. (extensions/msteams/src/feedback-reflection.ts:171, a6a4652c708a)

Likely related people:

SidU: The merged Teams AI UX PR added feedback reflection behavior and the original feedback reflection tests, including the prompt path this PR repairs. (role: introduced feature; confidence: high; commits: cd90130877f1, 15158d358acc; files: extensions/msteams/src/feedback-reflection.ts, extensions/msteams/src/feedback-reflection.test.ts)
jacobtomlinson: A later merged Teams feedback PR changed feedback invoke authorization around the same feedback/reflection workflow boundary. (role: recent adjacent contributor; confidence: medium; commits: c5415a474bb0, 1dede6d05fee; files: extensions/msteams/src/monitor-handler.ts, extensions/msteams/src/monitor-handler.feedback-authz.test.ts)
steipete: A recent plugin-state refactor touched Microsoft Teams feedback learning storage and feedback-reflection.test.ts, making them a useful routing candidate for adjacent feedback-state context. (role: recent adjacent contributor; confidence: medium; commits: 33c246dbbaaa, 26d5557c0e77; files: extensions/msteams/src/feedback-reflection-store.ts, extensions/msteams/src/feedback-reflection.test.ts)

What the crustacean ranks mean

🦀 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.

How this review workflow works

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.

buildReflectionPrompt truncated the thumbed-down response with String.slice(0, 500) on a UTF-16 code-unit index, so an astral character straddling the 500-char cap was cut into a lone surrogate in the reflection prompt built for the LLM. Use truncateUtf16Safe so truncation never splits a surrogate pair, keeping the existing 500-char budget and '...' suffix. Adds tests asserting the prompt stays UTF-16 well formed when truncating and that a boundary emoji is preserved when it fits.

…#96578) buildReflectionPrompt truncated the thumbed-down response with String.slice(0, 500) on a UTF-16 code-unit index, so an astral character straddling the 500-char cap was cut into a lone surrogate in the reflection prompt built for the LLM. Use truncateUtf16Safe so truncation never splits a surrogate pair, keeping the existing 500-char budget and '...' suffix. Adds tests asserting the prompt stays UTF-16 well formed when truncating and that a boundary emoji is preserved when it fits.

…#96578) buildReflectionPrompt truncated the thumbed-down response with String.slice(0, 500) on a UTF-16 code-unit index, so an astral character straddling the 500-char cap was cut into a lone surrogate in the reflection prompt built for the LLM. Use truncateUtf16Safe so truncation never splits a surrogate pair, keeping the existing 500-char budget and '...' suffix. Adds tests asserting the prompt stays UTF-16 well formed when truncating and that a boundary emoji is preserved when it fits. (cherry picked from commit 6c5a9fd)

…#96578) buildReflectionPrompt truncated the thumbed-down response with String.slice(0, 500) on a UTF-16 code-unit index, so an astral character straddling the 500-char cap was cut into a lone surrogate in the reflection prompt built for the LLM. Use truncateUtf16Safe so truncation never splits a surrogate pair, keeping the existing 500-char budget and '...' suffix. Adds tests asserting the prompt stays UTF-16 well formed when truncating and that a boundary emoji is preserved when it fits.

…void lone surrogates The Mattermost inbound handler in extensions/mattermost/src/mattermost/monitor.ts truncates the message body preview with bodyText.slice(0, 200), which can split a UTF-16 surrogate pair when an emoji (e.g. 🎉, 🦞) or other astral character lands at the 200-char boundary. The resulting lone surrogate half leaks into the verbose log preview and could propagate to downstream consumers that reject ill-formed strings. Replace the raw .slice(0, 200) with the shared truncateUtf16Safe helper (from openclaw/plugin-sdk/text-utility-runtime), which backs up one code unit when the cut would land inside a surrogate pair. This matches the same pattern already applied to the Slack, Discord, Telegram, IRC, Line, MS Teams, and Feishu channel previews (openclaw#96456, openclaw#96569, openclaw#96572, openclaw#96577, openclaw#96578, openclaw#97462, openclaw#97472, openclaw#98994). The preview is only used for the verbose inbound log line so the behavioral change is cosmetic: when the 200th code unit would split an emoji, the preview now includes one fewer code unit rather than emitting a broken surrogate.

openclaw-barnacle Bot added channel: msteams Channel integration: msteams size: XS triage: needs-pr-context Candidate: external PR body lacks required problem context or evidence. labels Jun 24, 2026

WeeLi-009 force-pushed the fix/msteams-reflection-truncate-surrogate branch from e8f027b to d4da0d3 Compare June 25, 2026 00:23

BradGroux mentioned this pull request Jun 25, 2026

WORKING: All Microsoft Issues and PRs (refresh) #74163

Draft

openclaw-barnacle Bot removed the triage: needs-pr-context Candidate: external PR body lacks required problem context or evidence. label Jun 25, 2026

github-actions Bot mentioned this pull request Jun 26, 2026

🦞 OpenClaw 生态日报 2026-06-26 96loveslife/big_model_radar#60

Open

WeeLi-009 force-pushed the fix/msteams-reflection-truncate-surrogate branch from d4da0d3 to 32f9331 Compare June 26, 2026 06:24

vincentkoc merged commit 6c5a9fd into openclaw:main Jun 27, 2026
87 checks passed

github-actions Bot mentioned this pull request Jun 27, 2026

📡 Upstream Digest — 2026-06-27 16:59 UTC curtismercier/openclaw-mods#1160

Open

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Uh oh!

fix(msteams): truncate reflection prompt on a UTF-16 boundary#96578

fix(msteams): truncate reflection prompt on a UTF-16 boundary#96578
vincentkoc merged 1 commit into
openclaw:mainfrom
WeeLi-009:fix/msteams-reflection-truncate-surrogate

WeeLi-009 commented Jun 24, 2026 •

edited

Loading

Uh oh!

clawsweeper Bot commented Jun 24, 2026 •

edited

Loading

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Uh oh!

Uh oh!

Conversation

WeeLi-009 commented Jun 24, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

What Problem This Solves

Fix

Evidence

Tests

Uh oh!

clawsweeper Bot commented Jun 24, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

WeeLi-009 commented Jun 24, 2026 •

edited

Loading

clawsweeper Bot commented Jun 24, 2026 •

edited

Loading