fix(active-memory): keep recall summaries UTF-16 safe at truncation boundary by ZengWen-DT · Pull Request #98711 · openclaw/openclaw

ZengWen-DT · 2026-07-01T16:23:34Z

What Problem This Solves

Fixes an issue where the active-memory recall summary that gets prepended to the agent prompt could contain a lone UTF-16 surrogate half when the summary was long enough to be truncated and an emoji (or other astral character) sat exactly on the truncation boundary. The broken half renders as a replacement character in the injected <active_memory_plugin> context.

Why This Change Was Made

truncateSummary now preserves the old whitespace/word-boundary decision from the raw cut, but uses the existing surrogate-safe sliceUtf16Safe helper for the returned text. That keeps the previous word-boundary behavior while ensuring the reserved-ellipsis cut can no longer split a surrogate pair. This mirrors the recently merged tool-summary fix (#98644), which applied the same helper to the same class of user-visible truncation.

User Impact

Recall summaries containing emoji or other astral Unicode now truncate cleanly before the split character instead of leaking a broken half-character into prompt context. Existing word-boundary truncation is preserved, including the whitespace-before-emoji boundary case raised in review.

Evidence

Premise: String.prototype.slice operates on UTF-16 code units, so slice(0, n) where index n falls inside a surrogate pair keeps a dangling high surrogate. sliceUtf16Safe (openclaw/plugin-sdk/text-utility-runtime) is the existing accepted helper for this exact case and is already used by the merged tool-summary fix (#98644) and by sibling channels (Discord model picker, Feishu streaming card).

Review finding fixed: added coverage for the whitespace-before-emoji boundary. The regression drives the real before_prompt_build active-memory hook and verifies alpha beta cccccccccccccccccccccccccc… is kept instead of falling back to alpha beta…, while TAILWORD is not included.

node scripts/run-vitest.mjs run extensions/active-memory/index.test.ts -t "truncates recall summaries|keeps the old word boundary"
Test Files  1 passed (1)
Tests  2 passed | 154 skipped (156)

Full focused active-memory test file after the revision:

node scripts/run-vitest.mjs run extensions/active-memory/index.test.ts
Test Files  1 passed (1)
Tests  156 passed (156)

Real OpenClaw run proof: ran pnpm openclaw agent --local --session-key agent:main:direct:proof3 --channel webchat --message "what should I remember about the emoji boundary?" --verbose on --json --timeout 60 with an isolated temporary OPENCLAW_STATE_DIR, active-memory enabled, maxSummaryChars: 40, and a local mock OpenAI-compatible provider. The recall subagent made a real memory_get tool call against a temporary MEMORY.md; the mock provider then returned the emoji-bearing recall summary alpha beta cccccccccccccccccccccccccc 🎉TAILWORD. The main model request contained the hidden active-memory prompt prefix after truncation:

request entries: 3
recall tool call requested memory_get: true
recall tool output present: true
main prompt prefix: alpha beta cccccccccccccccccccccccccc…
main prefix has TAILWORD: false
main prefix has replacement char: false
main prefix length: 38

Type-check, format, and lint on the changed files:

pnpm format:check -- extensions/active-memory/index.ts extensions/active-memory/index.test.ts
All matched files use the correct format.

OPENCLAW_OXLINT_SKIP_PREPARE=1 node scripts/run-oxlint.mjs extensions/active-memory/index.ts extensions/active-memory/index.test.ts
exit 0

pnpm tsgo:extensions:test
exit 0

What was not tested: no external live provider credentials were used. The after-fix proof used OpenClaw's real local agent runtime, real active-memory plugin loading, a real active-memory recall subagent turn, and a local OpenAI-compatible test provider so the prompt injection behavior could be captured without sending private proof data to an external model.

AI-assisted: implemented with Cursor; all proof commands were run manually.

…oundary Co-authored-by: Cursor <[email protected]>

clawsweeper · 2026-07-01T16:26:13Z

Codex review: needs maintainer review before merge. Reviewed July 2, 2026, 10:06 PM ET / 02:06 UTC.

Summary
The PR imports sliceUtf16Safe into active-memory summary truncation and adds hook-level regression tests for surrogate-boundary and whitespace-before-emoji truncation cases.

PR surface: Source +3, Tests +72. Total +75 across 2 files.

Reproducibility: yes. for source-level reproduction: current main slices active-memory recall summaries by raw UTF-16 code units, and the read-only probe shows a dangling surrogate when an emoji straddles the cut. I did not run the full OpenClaw hook locally because this review is read-only, but the PR adds hook-level regression coverage for the path.

Review metrics: none identified.

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

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

Next step before merge

No ClawSweeper repair job is needed; the remaining action is ordinary maintainer review and merge decision for the contributor branch.

Security
Cleared: The diff only changes active-memory text truncation and tests, imports an existing SDK helper, and does not alter dependencies, package scripts, secrets, CI, or privileged runtime behavior.

Review details

Best possible solution:

Keep this PR open until a linked canonical PR proves it covers this PR's unique work, or a maintainer confirms closure.

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

Yes for source-level reproduction: current main slices active-memory recall summaries by raw UTF-16 code units, and the read-only probe shows a dangling surrogate when an emoji straddles the cut. I did not run the full OpenClaw hook locally because this review is read-only, but the PR adds hook-level regression coverage for the path.

Is this the best way to solve the issue?

Yes: the final patch keeps the old raw-cut word-boundary decision and uses sliceUtf16Safe only for returned text, which fixes the UTF-16 boundary bug without adding a new config surface or duplicating a generic truncation helper.

AGENTS.md: found and applied where relevant.

Codex review notes: model internal, reasoning high; reviewed against 3ad465d32b3b.

Label changes

Label changes:

add proof: sufficient: Contributor real behavior proof is sufficient. The PR body now includes copied terminal output from an after-fix OpenClaw agent run with active-memory enabled, a temporary state dir, a real recall subagent turn, and captured prompt-prefix output showing no replacement character or leaked tail word.
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 (terminal): The PR body now includes copied terminal output from an after-fix OpenClaw agent run with active-memory enabled, a temporary state dir, a real recall subagent turn, and captured prompt-prefix output showing no replacement character or leaked tail word.
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:

P3: This is a narrow active-memory text-truncation correctness fix with limited blast radius and no evidence of urgent workflow breakage.
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 (terminal): The PR body now includes copied terminal output from an after-fix OpenClaw agent run with active-memory enabled, a temporary state dir, a real recall subagent turn, and captured prompt-prefix output showing no replacement character or leaked tail word.
proof: sufficient: Contributor real behavior proof is sufficient. The PR body now includes copied terminal output from an after-fix OpenClaw agent run with active-memory enabled, a temporary state dir, a real recall subagent turn, and captured prompt-prefix output showing no replacement character or leaked tail word.

Evidence reviewed

PR surface:

Source +3, Tests +72. Total +75 across 2 files.

View PR surface stats

Area	Files	Added	Removed	Net
Source	1	6	3	+3
Tests	1	72	0	+72
Docs	0	0	0	0
Config	0	0	0	0
Generated	0	0	0	0
Other	0	0	0	0
Total	2	78	3	+75

What I checked:

PR close coverage proof: PR close coverage proof kept this PR open against fix: tool summaries preserve emoji truncation boundaries #98644: PR B is a merged sibling fix for the same bug class, but it does not carry PR A's material active-memory recall summary behavior. The source report identifies remaining current-main work in the active-memory path, so PR A should not be closed as covered by PR B.
linked superseding PR: fix: tool summaries preserve emoji truncation boundaries #98644 (fix: tool summaries preserve emoji truncation boundaries) is merged at 2026-07-01T15:56:41Z.
cluster evidence: the durable review links that PR in the work cluster or recommended risk path.
no human follow-up: live comments and timeline hydrated by apply contain no non-automation activity after the ClawSweeper review.

Likely related people:

Takhoffman: Authored and merged the Active Memory recall plugin and the later move of recall summaries into the hidden prompt prefix, which is the runtime surface this PR changes. (role: feature owner; confidence: high; commits: b83726d13e33, f94d6778b1d5; files: extensions/active-memory/index.ts, extensions/active-memory/index.test.ts)
ZengWen-DT: Authored the merged sibling UTF-16 truncation fix for tool-description summaries, the same bug class this active-memory PR mirrors. (role: recent adjacent contributor; confidence: medium; commits: 85572835ae46, d9d6108086da; files: src/agents/tool-description-summary.ts, src/agents/tool-description-summary.test.ts)
vincentkoc: Merged the related UTF-16-safe tool-summary PR that established the nearby accepted helper-based fix pattern. (role: merger of adjacent fix; confidence: medium; commits: d9d6108086da; files: src/agents/tool-description-summary.ts, src/agents/tool-description-summary.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.

ZengWen-DT · 2026-07-02T01:53:17Z

@clawsweeper re-review

clawsweeper · 2026-07-02T01:53:20Z

🦞🧹
ClawSweeper re-review requested.

I asked ClawSweeper to review this item again.
Action: item re-review queued (workflow sweep.yml, event repository_dispatch).
Result: the existing ClawSweeper review comment will be edited in place when the review finishes.

ZengWen-DT · 2026-07-03T01:49:16Z

@clawsweeper re-review

vincentkoc · 2026-07-05T12:49:17Z

Thanks @ZengWen-DT. #100244 folded this active-memory repair into the canonical backend UTF-16 boundary fix and preserved your co-author credit.

Landed on main in 66081c0, including the active-memory implementation and focused regression coverage. Exact-head CI passed at https://github.com/openclaw/openclaw/actions/runs/28741038571.

Closing this PR as superseded. A concrete active-memory truncation case not covered by the landed implementation would be grounds for a new issue or reopening.

fix(active-memory): keep recall summaries UTF-16 safe at truncation b…

89592d9

…oundary Co-authored-by: Cursor <[email protected]>

openclaw-barnacle Bot added the size: XS label Jul 1, 2026

fix(active-memory): preserve recall word boundary

460e04d

openclaw-barnacle Bot added size: S and removed size: XS labels Jul 2, 2026

vincentkoc mentioned this pull request Jul 5, 2026

fix(core): keep backend truncation UTF-16 safe #100244

Merged

vincentkoc closed this Jul 5, 2026

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Uh oh!

fix(active-memory): keep recall summaries UTF-16 safe at truncation boundary#98711

fix(active-memory): keep recall summaries UTF-16 safe at truncation boundary#98711
ZengWen-DT wants to merge 2 commits into
openclaw:mainfrom
ZengWen-DT:fix/active-memory-summary-utf16

ZengWen-DT commented Jul 1, 2026 •

edited

Loading

Uh oh!

clawsweeper Bot commented Jul 1, 2026 •

edited

Loading

Uh oh!

ZengWen-DT commented Jul 2, 2026

Uh oh!

clawsweeper Bot commented Jul 2, 2026 •

edited

Loading

Uh oh!

ZengWen-DT commented Jul 3, 2026

Uh oh!

vincentkoc commented Jul 5, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Uh oh!

Uh oh!

Conversation

ZengWen-DT commented Jul 1, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

What Problem This Solves

Why This Change Was Made

User Impact

Evidence

Uh oh!

clawsweeper Bot commented Jul 1, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

ZengWen-DT commented Jul 2, 2026

Uh oh!

clawsweeper Bot commented Jul 2, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

ZengWen-DT commented Jul 3, 2026

Uh oh!

vincentkoc commented Jul 5, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

ZengWen-DT commented Jul 1, 2026 •

edited

Loading

clawsweeper Bot commented Jul 1, 2026 •

edited

Loading

clawsweeper Bot commented Jul 2, 2026 •

edited

Loading