fix: guard against malformed text blocks in context truncation by cgdusek · Pull Request #51267 · openclaw/openclaw

cgdusek · 2026-03-20T20:47:34Z

Summary

Guard isTextBlock type guard to require typeof block.text === "string" (not just block.type === "text")
Guard 3 additional block.type === "text" sites in pruner.ts (collectTextSegments, collectPrunableToolResultSegments, estimateTextAndImageChars)
Add regression tests covering malformed {type: "text"} blocks in both the char estimator and context pruner

A plugin tool handler returning undefined/void produces a malformed content block {type: "text"} (no text property). This block persists to session JSONL and crashes the context truncation pipeline with TypeError: Cannot read properties of undefined (reading 'length') on every subsequent message, locking the session into an unrecoverable crash loop.

Closes #34979

Test plan

pnpm test -- src/agents/pi-embedded-runner/tool-result-char-estimator.test.ts passes (4 new tests)
pnpm test -- src/agents/pi-extensions/context-pruning/pruner.test.ts passes (2 new + 11 existing tests)
CI green

Based on prior work by @alvinttang (#39331) and @coffeexcoin (#34980).

🤖 Generated with Claude Code

greptile-apps · 2026-03-20T20:51:34Z

Greptile Summary

This PR hardens the context-truncation pipeline against a class of malformed content blocks produced when a plugin tool handler returns undefined/void. Those handlers write {type: "text"} (no text property) to the session JSONL, which previously caused TypeError: Cannot read properties of undefined (reading 'length') on every subsequent message — creating an unrecoverable crash loop.

Key changes:

isTextBlock type guard in tool-result-char-estimator.ts now requires typeof block.text === "string" in addition to type === "text", correctly routing malformed blocks to the estimateUnknownChars fallback.
Three additional guard sites in pruner.ts (collectTextSegments, collectPrunableToolResultSegments, estimateTextAndImageChars) add the same typeof block.text === "string" check, silently skipping malformed blocks rather than crashing.
Regression tests are added for both files, covering the crash paths described in issue [Bug]: Malformed tool result content block crashes context truncation in a session-permanent loop #34979.

The logic changes are minimal, consistent, and defensive without altering any behavior for well-formed blocks. One minor test assertion in tool-result-char-estimator.test.ts is weaker than its comment intends (toBeGreaterThanOrEqual(0) instead of toBeGreaterThan(0)).

Confidence Score: 5/5

This PR is safe to merge — it adds narrowly-scoped defensive guards with no behavioral change for well-formed blocks.
The fix is minimal and surgical: three one-line guard additions in pruner.ts and one condition added to an existing type guard in tool-result-char-estimator.ts. All existing tests continue to pass, and new regression tests confirm the crash is resolved. There is no change to control flow for valid data.
No files require special attention.

Prompt To Fix All With AI

This is a comment left during a code review.
Path: src/agents/pi-embedded-runner/tool-result-char-estimator.test.ts
Line: 30

Comment:
**Test assertion doesn't match comment intent**

The comment says the malformed block should be estimated "not zero" (i.e. the `estimateUnknownChars` fallback should fire), but `toBeGreaterThanOrEqual(0)` allows a result of `0` — which would silently pass even if the fallback were broken and returned nothing. The assertion should be `toBeGreaterThan(0)` to actually enforce the stated invariant.

For reference, `estimateUnknownChars({type:"text"})` will `JSON.stringify` the object to `'{"type":"text"}'` (15 chars), so the expected result is already well above zero.

```suggestion
    expect(estimateMessageCharsCached(malformed, cache)).toBeGreaterThan(0);
```

How can I resolve this? If you propose a fix, please make it concise.

_{Last reviewed commit: "fix: guard against m..."}

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: a0a2700b2e

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

cgdusek · 2026-03-20T20:55:52Z

Addressed both review comments in b5cd499:

Greptile (P2) — test assertion: Tightened from toBeGreaterThanOrEqual(0) to toBeGreaterThan(0). The fallback estimateUnknownChars JSON-stringifies {type:"text"} to 15 chars, so this now actually enforces the invariant.

Codex (P1) — null content entries: Added if (!block) continue; guards to all four content block loops in pruner.ts (collectTextSegments, collectPrunableToolResultSegments, hasImageBlocks, estimateTextAndImageChars). Added a regression test for toolResult with null content entries.

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: adca2808f3

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

cgdusek · 2026-03-20T21:17:52Z

Re: Codex P1 — Preserve size accounting for malformed text blocks

Valid defense-in-depth observation, but out of scope for this crash fix:

Before this PR: {type: "text"} crashes block.text.length → session permanently locked in a crash loop (the reported bug).

After this PR: malformed block is silently skipped in the pruner's size estimate → minor undercount, but the session is functional and recoverable.

The char estimator (tool-result-char-estimator.ts:54-66) already handles this correctly via its estimateUnknownChars fallback (JSON.stringify). Adding equivalent fallback logic to the pruner's estimateTextAndImageChars would require broadening the pruner's type contract (ReadonlyArray<TextContent | ImageContent> → unknown[]). The crash → silent-skip is a strict improvement for the reported issue. Happy to do that as a follow-up if desired.

cgdusek · 2026-03-20T21:47:00Z

CI status — remaining failures are pre-existing / unrelated

All PR-scoped checks pass (lint, build, unit tests). The 3 remaining red jobs are unrelated to this change:

Job	Failing tests	Touched by this PR?	On main?
`checks-windows (node, test, 2, 6)`	`src/node-host/invoke.sanitize-env.test.ts` — Windows env key handling	No	Yes — confirmed failing on main (run 23362337869)
`checks (node, channels)`	`extensions/telegram/src/send.proxy.test.ts`, `extensions/whatsapp/src/inbound.media.test.ts`	No	Flaky on main
`checks (node, extensions)`	`extensions/tavily/index.test.ts`, `extensions/matrix/index.test.ts`, `extensions/bluebubbles/src/send.test.ts`	No	Flaky on main

This PR only modifies:

src/agents/pi-embedded-runner/tool-result-char-estimator.ts
src/agents/pi-embedded-runner/tool-result-char-estimator.test.ts (new)
src/agents/pi-extensions/context-pruning/pruner.ts
src/agents/pi-extensions/context-pruning/pruner.test.ts

None of the failing tests are in these files or import from them.

alberthild · 2026-03-23T09:15:39Z

We audited this comprehensively on our end. The type guards are placed exactly where they need to be, handling both malformed objects and null entries correctly across both the estimator and pruner pipelines. The regression tests cover the crash edge cases perfectly. Fully support merging this to kill the crash loop! ✅

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 87d6b79b55

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

aisle-research-bot · 2026-03-23T22:51:27Z

🔒 Aisle Security Analysis

We found 2 potential security issue(s) in this PR:

#	Severity	Title
1	🟠 High	Potential DoS via unbounded JSON.stringify of malformed tool-result text blocks
2	🟡 Medium	Malformed toolResult text blocks are JSON-stringified and injected into pruned LLM context

1. 🟠 Potential DoS via unbounded JSON.stringify of malformed tool-result text blocks

Property	Value
Severity	High
CWE	CWE-400
Location	`src/agents/pi-hooks/context-pruning/pruner.ts:16-34`

Description

pruneContextMessages() now attempts to handle malformed {type:"text"} blocks by serializing the entire block with JSON.stringify() when text is not a string.

Input/control: tool results can originate from tools/plugins (potentially 3rd-party or remote) and are stored in the session transcript.
Issue: if a tool returns a text block where text is an object/array (possibly huge or deeply nested), serializeMalformedTextBlock() will call JSON.stringify(block) with no size/depth limits.
Impact: JSON.stringify() on large/deep inputs can cause high CPU usage and memory pressure, and the resulting large string is then further processed by character estimation and trimming logic (amplifying work). Because context pruning runs during context assembly, this can be triggered repeatedly, leading to request-time resource exhaustion.

Vulnerable code:

function serializeMalformedTextBlock(block: unknown): string {
  try {
    const serialized = JSON.stringify(block);
    return typeof serialized === "string" ? serialized : "[malformed text block]";
  } catch {
    return "[malformed text block]";
  }
}

function coerceTextBlock(block: unknown): string | null {
  ...
  return typeof text === "string" ? text : serializeMalformedTextBlock(block);
}

Recommendation

Avoid serializing arbitrary malformed blocks without bounds.

Recommended options (prefer 1 or 2):

Do not stringify malformed blocks at all; replace with a constant marker.

function coerceTextBlock(block: unknown): string | null {
  if (!block || typeof block !== "object") return null;
  if ((block as any).type !== "text") return null;
  const text = (block as any).text;
  return typeof text === "string" ? text : "[malformed text block]";
}

If you need some diagnostics, use a bounded safe stringify (max depth / max output length), and never stringify the whole block if it may contain large fields.

import { safeStringify } from "./safe-stringify"; // implement depth/length caps

function serializeMalformedTextBlock(block: unknown): string {
  return safeStringify(block, { maxDepth: 4, maxLength: 2048 }) ?? "[malformed text block]";
}

Additionally, consider caching the coerced text per block/message during a single pruning run to avoid repeated serialization across estimate* and collect* passes.

2. 🟡 Malformed toolResult text blocks are JSON-stringified and injected into pruned LLM context

Property	Value
Severity	Medium
CWE	CWE-201
Location	`src/agents/pi-hooks/context-pruning/pruner.ts:16-34`

Description

The context pruner now attempts to recover from malformed {type:"text"} blocks by serializing the entire block with JSON.stringify() and treating that serialization as user-visible text.

This can lead to unintended disclosure and prompt-injection amplification:

Input: toolResult.content[] blocks can originate from plugins/tools and may be malformed or attacker-influenced.
Transformation: when block.type === "text" but block.text is not a string, coerceTextBlock() falls back to serializeMalformedTextBlock(block).
Sink: the resulting JSON string is included in collectTextSegments() / collectPrunableToolResultSegments() and may be embedded into the pruned context sent to the LLM (e.g., during image-path normalization or trimming).
Impact: JSON serialization can expose extra fields that were not intended to be model-visible (e.g., internal metadata, tokens, file contents embedded in structured objects) and can carry structured prompt-injection payloads verbatim.

Vulnerable code:

function serializeMalformedTextBlock(block: unknown): string {
  try {
    const serialized = JSON.stringify(block);
    return typeof serialized === "string" ? serialized : "[malformed text block]";
  } catch {
    return "[malformed text block]";
  }
}

function coerceTextBlock(block: unknown): string | null {
  ...
  const text = (block as { text?: unknown }).text;
  return typeof text === "string" ? text : serializeMalformedTextBlock(block);
}

Recommendation

Avoid serializing arbitrary malformed blocks into LLM-visible context.

Recommended approaches:

Replace with a fixed placeholder (safe-by-default) and optionally include only minimal diagnostics:

function coerceTextBlock(block: unknown): string | null {
  if (!block || typeof block !== "object") return null;
  if ((block as any).type !== "text") return null;
  const text = (block as any).text;
  return typeof text === "string" ? text : "[malformed text block removed]";
}

If you must preserve some information, allowlist specific fields and cap length:

const safe = { type: (block as any).type };
return JSON.stringify(safe).slice(0, 256);

Additionally, consider applying existing secret-redaction logic (if any) to any fallback text before inserting it into the model context.

Analyzed PR: #51267 at commit ca7d630

_{Last updated on: 2026-04-25T06:32:29Z}

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: cfeebbcbc3

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

chatgpt-codex-connector · 2026-03-23T22:55:53Z

+    dirName: "google-gemini-cli-auth",
+    idHint: "google-gemini-cli-auth",
+    source: {
+      source: "./index.ts",
+      built: "index.js",


Remove nonexistent auth plugins from bundled metadata

collectBundledPluginMetadata() in scripts/generate-bundled-plugin-metadata.mjs only serializes extensions/* directories that still have both package.json and openclaw.plugin.json, and this commit contains neither extensions/google-gemini-cli-auth nor extensions/minimax-portal-auth. Because these snapshot entries were added anyway, regenerating src/plugins/bundled-plugin-metadata.generated.ts drops them again (I verified the collector returns 75 entries while the checked-in snapshot now has 77), so the matches the generated metadata snapshot check in src/plugins/bundled-plugin-metadata.test.ts will fail and CI will keep reporting the generated file as stale.

Useful? React with 👍 / 👎.

@alvinttang

…ar estimation A plugin tool handler returning undefined produces {type: "text"} (no text property) in the session JSONL. This crashes the context truncation pipeline with TypeError on every subsequent message, locking the session into an unrecoverable crash loop. Guard all four crash sites: - isTextBlock type guard in tool-result-char-estimator.ts - collectTextSegments in pruner.ts - collectPrunableToolResultSegments in pruner.ts - estimateTextAndImageChars in pruner.ts Add regression tests for both the char estimator and the context pruner covering malformed toolResult content blocks. Closes openclaw#34979 Based on prior work by @alvinttang (openclaw#39331) and @coffeexcoin (openclaw#34980). Co-Authored-By: alvinttang <alvinttang@users.noreply.github.com> Co-Authored-By: coffeexcoin <coffeexcoin@users.noreply.github.com> Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

- Tighten test assertion from toBeGreaterThanOrEqual(0) to toBeGreaterThan(0) so the unknown-block fallback is actually enforced (greptile review). - Add null/undefined guards to all content block loops in pruner.ts to prevent TypeError on block.type when sessions contain null entries (codex review). - Add regression test for toolResult with null content entries in pruner.test.ts. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

BunsDev · 2026-04-25T06:30:58Z

Maintainer update before landing:

Rebased onto current main after the context-pruning path moved from pi-extensions to pi-hooks.
Dropped the obsolete generated plugin metadata churn from the old branch.
Addressed the Aisle undercount finding: malformed non-string text blocks are now serialized for pruning/size accounting instead of being treated as zero-sized.
Added a regression test proving malformed non-string text blocks trigger tool-result trimming.

Local verification:

pnpm test src/agents/pi-embedded-runner/tool-result-char-estimator.test.ts src/agents/pi-hooks/context-pruning/pruner.test.ts
pnpm check:changed

Waiting for the fresh GitHub checks before merge.

@ajfonthemove

Three external-contributor commits from the last day landed without CHANGELOG entries: - Alex Fries (#68286, @ajfonthemove): hybrid memory search component scores. Added under Unreleased > Changes (feat). - Charles Dusek (#51267, @cgdusek): malformed tool-result text-block guard. Added under Unreleased > Fixes. - Jerome Benoit (#59935, @jerome-benoit): Nix Home Manager daemon PATH support. Added under Unreleased > Fixes. Also drop a duplicate raw-subject changelog line for #66884 that restated alexlomt's already-formatted entry one line above.

@vincentkoc

… tool-result pruning entries Three entries were missing co-credits I should have preserved: - Diagnostics/OTEL exec-process spans (#71451): @vincentkoc implemented, but @jlapenna's #70424 proposed the broader tracing work this entry builds on. Now credits both. - Diagnostics/OTEL preloaded SDK (#71450): same pattern — credits @vincentkoc and @jlapenna. - Agents/tool-result pruning (#51267): @cgdusek's PR explicitly built on prior work in #39331 by @alvinttang and #34980 by @coffeexcoin. Now credits all three.

@ajfonthemove

Three external-contributor commits from the last day landed without CHANGELOG entries: - Alex Fries (openclaw#68286, @ajfonthemove): hybrid memory search component scores. Added under Unreleased > Changes (feat). - Charles Dusek (openclaw#51267, @cgdusek): malformed tool-result text-block guard. Added under Unreleased > Fixes. - Jerome Benoit (openclaw#59935, @jerome-benoit): Nix Home Manager daemon PATH support. Added under Unreleased > Fixes. Also drop a duplicate raw-subject changelog line for openclaw#66884 that restated alexlomt's already-formatted entry one line above.

@vincentkoc

… tool-result pruning entries Three entries were missing co-credits I should have preserved: - Diagnostics/OTEL exec-process spans (openclaw#71451): @vincentkoc implemented, but @jlapenna's openclaw#70424 proposed the broader tracing work this entry builds on. Now credits both. - Diagnostics/OTEL preloaded SDK (openclaw#71450): same pattern — credits @vincentkoc and @jlapenna. - Agents/tool-result pruning (openclaw#51267): @cgdusek's PR explicitly built on prior work in openclaw#39331 by @alvinttang and openclaw#34980 by @coffeexcoin. Now credits all three.

@ajfonthemove

Three external-contributor commits from the last day landed without CHANGELOG entries: - Alex Fries (openclaw#68286, @ajfonthemove): hybrid memory search component scores. Added under Unreleased > Changes (feat). - Charles Dusek (openclaw#51267, @cgdusek): malformed tool-result text-block guard. Added under Unreleased > Fixes. - Jerome Benoit (openclaw#59935, @jerome-benoit): Nix Home Manager daemon PATH support. Added under Unreleased > Fixes. Also drop a duplicate raw-subject changelog line for openclaw#66884 that restated alexlomt's already-formatted entry one line above.

@vincentkoc

… tool-result pruning entries Three entries were missing co-credits I should have preserved: - Diagnostics/OTEL exec-process spans (openclaw#71451): @vincentkoc implemented, but @jlapenna's openclaw#70424 proposed the broader tracing work this entry builds on. Now credits both. - Diagnostics/OTEL preloaded SDK (openclaw#71450): same pattern — credits @vincentkoc and @jlapenna. - Agents/tool-result pruning (openclaw#51267): @cgdusek's PR explicitly built on prior work in openclaw#39331 by @alvinttang and openclaw#34980 by @coffeexcoin. Now credits all three.

@ajfonthemove

Three external-contributor commits from the last day landed without CHANGELOG entries: - Alex Fries (openclaw#68286, @ajfonthemove): hybrid memory search component scores. Added under Unreleased > Changes (feat). - Charles Dusek (openclaw#51267, @cgdusek): malformed tool-result text-block guard. Added under Unreleased > Fixes. - Jerome Benoit (openclaw#59935, @jerome-benoit): Nix Home Manager daemon PATH support. Added under Unreleased > Fixes. Also drop a duplicate raw-subject changelog line for openclaw#66884 that restated alexlomt's already-formatted entry one line above.

@vincentkoc

… tool-result pruning entries Three entries were missing co-credits I should have preserved: - Diagnostics/OTEL exec-process spans (openclaw#71451): @vincentkoc implemented, but @jlapenna's openclaw#70424 proposed the broader tracing work this entry builds on. Now credits both. - Diagnostics/OTEL preloaded SDK (openclaw#71450): same pattern — credits @vincentkoc and @jlapenna. - Agents/tool-result pruning (openclaw#51267): @cgdusek's PR explicitly built on prior work in openclaw#39331 by @alvinttang and openclaw#34980 by @coffeexcoin. Now credits all three.

@ajfonthemove

Three external-contributor commits from the last day landed without CHANGELOG entries: - Alex Fries (openclaw#68286, @ajfonthemove): hybrid memory search component scores. Added under Unreleased > Changes (feat). - Charles Dusek (openclaw#51267, @cgdusek): malformed tool-result text-block guard. Added under Unreleased > Fixes. - Jerome Benoit (openclaw#59935, @jerome-benoit): Nix Home Manager daemon PATH support. Added under Unreleased > Fixes. Also drop a duplicate raw-subject changelog line for openclaw#66884 that restated alexlomt's already-formatted entry one line above.

@vincentkoc

… tool-result pruning entries Three entries were missing co-credits I should have preserved: - Diagnostics/OTEL exec-process spans (openclaw#71451): @vincentkoc implemented, but @jlapenna's openclaw#70424 proposed the broader tracing work this entry builds on. Now credits both. - Diagnostics/OTEL preloaded SDK (openclaw#71450): same pattern — credits @vincentkoc and @jlapenna. - Agents/tool-result pruning (openclaw#51267): @cgdusek's PR explicitly built on prior work in openclaw#39331 by @alvinttang and openclaw#34980 by @coffeexcoin. Now credits all three.

@ajfonthemove

Three external-contributor commits from the last day landed without CHANGELOG entries: - Alex Fries (openclaw#68286, @ajfonthemove): hybrid memory search component scores. Added under Unreleased > Changes (feat). - Charles Dusek (openclaw#51267, @cgdusek): malformed tool-result text-block guard. Added under Unreleased > Fixes. - Jerome Benoit (openclaw#59935, @jerome-benoit): Nix Home Manager daemon PATH support. Added under Unreleased > Fixes. Also drop a duplicate raw-subject changelog line for openclaw#66884 that restated alexlomt's already-formatted entry one line above.

@vincentkoc

… tool-result pruning entries Three entries were missing co-credits I should have preserved: - Diagnostics/OTEL exec-process spans (openclaw#71451): @vincentkoc implemented, but @jlapenna's openclaw#70424 proposed the broader tracing work this entry builds on. Now credits both. - Diagnostics/OTEL preloaded SDK (openclaw#71450): same pattern — credits @vincentkoc and @jlapenna. - Agents/tool-result pruning (openclaw#51267): @cgdusek's PR explicitly built on prior work in openclaw#39331 by @alvinttang and openclaw#34980 by @coffeexcoin. Now credits all three.

greptile-apps Bot reviewed Mar 20, 2026

View reviewed changes

Comment thread src/agents/pi-embedded-runner/tool-result-char-estimator.test.ts Outdated

chatgpt-codex-connector Bot reviewed Mar 20, 2026

View reviewed changes

Comment thread src/agents/pi-extensions/context-pruning/pruner.ts Outdated

openclaw-barnacle Bot added agents Agent runtime and tooling size: S labels Mar 20, 2026

chatgpt-codex-connector Bot reviewed Mar 20, 2026

View reviewed changes

Comment thread src/agents/pi-extensions/context-pruning/pruner.ts Outdated

cgdusek mentioned this pull request Mar 20, 2026

[Bug]: Malformed tool result content block crashes context truncation in a session-permanent loop #34979

Closed

alberthild mentioned this pull request Mar 23, 2026

fix: guard estimateMessageCharsCached against null/undefined messages #41178

Closed

5 tasks

cgdusek force-pushed the claude/practical-wescoff branch from adca280 to 87d6b79 Compare March 23, 2026 22:17

chatgpt-codex-connector Bot reviewed Mar 23, 2026

View reviewed changes

Comment thread src/agents/pi-embedded-runner/tool-result-char-estimator.ts

openclaw-barnacle Bot added size: M and removed size: S labels Mar 23, 2026

chatgpt-codex-connector Bot reviewed Mar 23, 2026

View reviewed changes

openclaw-barnacle Bot added size: S and removed size: M labels Mar 23, 2026

cgdusek and others added 4 commits April 25, 2026 01:26

style: add braces to if-continue guards (oxlint curly rule)

3613be7

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

fix: count malformed text blocks during pruning

ca7d630

BunsDev force-pushed the claude/practical-wescoff branch from 60af534 to ca7d630 Compare April 25, 2026 06:30

openclaw-barnacle Bot added size: M and removed size: S labels Apr 25, 2026

BunsDev merged commit 6b38714 into openclaw:main Apr 25, 2026
66 checks passed

BunsDev mentioned this pull request Apr 25, 2026

fix: guard against malformed text blocks in context truncation #39331

Closed

3 tasks

Uh oh!

Conversation

cgdusek commented Mar 20, 2026

Summary

Test plan

Uh oh!

greptile-apps Bot commented Mar 20, 2026

Greptile Summary

Confidence Score: 5/5

Uh oh!

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

Uh oh!

cgdusek commented Mar 20, 2026

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

Uh oh!

cgdusek commented Mar 20, 2026

Uh oh!

cgdusek commented Mar 20, 2026

CI status — remaining failures are pre-existing / unrelated

Uh oh!

alberthild commented Mar 23, 2026

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

Uh oh!

aisle-research-bot Bot commented Mar 23, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

🔒 Aisle Security Analysis

Description

Recommendation

Description

Recommendation

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

chatgpt-codex-connector Bot Mar 23, 2026

Choose a reason for hiding this comment

Uh oh!

BunsDev commented Apr 25, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

aisle-research-bot Bot commented Mar 23, 2026 •

edited

Loading