MMSkillRisk: Can Agents Stay Safe When Multimodal Skills Become Traps?

Agent skills are shareable packages of procedural instructions, tools, and examples. Multimodal skills additionally include visual references that agents retrieve and inspect during execution. Because these images guide actions, attackers can disguise malicious instructions as ordinary visual...

Know the Normal, Track the Attack: Context-Grounded and Stateful LLM Investigation over System Provenance

Provenance-based intrusion detection systems PIDSs identify suspicious activity in audit streams, but their outputs remain difficult to turn into coherent attack narratives. Direct LLM analyses of local anomalous subgraphs lack deployment-specific normal-behavior knowledge and validated attack...

CVE-2026-101056 Cloudreve before 4.16.1 Authentication Bypass via Cached Context Hint

Cloudreve before 4.16.1 fails to revalidate share access when restoring cached navigator state from a contexthint UUID. Attackers who previously had valid share access can replay the cached hint to generate signed file URLs for up to 300 seconds after the share is deleted, expires, or reaches zer...

CVE-2026-100721 vm2 before 3.12.2 Authorization Bypass via Custom Resolver

vm2 before 3.12.2 contains an authorization bypass in the NodeVM external-module resolver. When an embedder configures require.external with a custom resolver and context: 'host', LegacyResolver.customResolve in lib/resolver-compat.js records the resolved module directory in this.externals as new...

When Consent Outlives Context: Residual Authority Replay in Long-Lived Agents

LLM agents increasingly rely on user approval to authorize security-sensitive actions at runtime. Such approvals are granted within a specific task and execution context. In long-lived agents, authorization decisions may need to persist across tasks or sessions. We find that this continuity can...

CVE-2026-100707 Kyverno before 1.19.1 Namespace Isolation Bypass via Percent-Encoded Path

Kyverno before 1.19.1 contains a namespace isolation bypass in the apiCall context entry of namespaced Policy resources due to inconsistent path interpretation between validation and execution. A low-privilege tenant can use percent-encoded dot-segments in urlPath to bypass namespace checks and...

CVE-2026-100707

Versions of Kyverno prior to 1.19.1 are vulnerable to a namespace isolation bypass within the apiCall context entry of namespaced Policy resources. The root cause is an inconsistent path interpretation between the validation and execution phases. A low-privilege tenant can exploit this by using p...

Astra Linux – Vulnerability found in Linux 6.1, Linux 6.12

In the Linux kernel, the following vulnerability has been resolved: perf/core: Detaching event groups during removeonexec The perfeventremoveonexec function removes events by calling perfeventexitevent. For top-level events, this removes the event from the context with DETACHEXIT only. This can...

CVE-2026-100553

OpenClaw versions = 2026.6.9 and 2026.8.1 do not declare the native chatId parameter as a delivery target in the Feishu unpin feature, so unpin requests can bypass the shared same-provider cross-context target check. When tools.message.crossContext.allowWithinProvider is disabled, an admitted...

CVE-2026-100530

OpenClaw versions before 2026.8.1 fail to bind working directory context to reusable exec approvals, allowing approved commands to execute in different directories. Attackers with an allow-always approval can reuse it to run the same command against unreviewed files or repositories with materiall...

CVE-2026-100553 OpenClaw 2026.6.9 before 2026.8.1 Cross-Context Policy Bypass via Feishu unpin

OpenClaw versions = 2026.6.9 and 2026.8.1 do not declare the native chatId parameter as a delivery target in the Feishu unpin feature, so unpin requests can bypass the shared same-provider cross-context target check. When tools.message.crossContext.allowWithinProvider is disabled, an admitted...

CVE-2026-100553 OpenClaw 2026.6.9 before 2026.8.1 Cross-Context Policy Bypass via Feishu unpin

OpenClaw versions = 2026.6.9 and 2026.8.1 do not declare the native chatId parameter as a delivery target in the Feishu unpin feature, so unpin requests can bypass the shared same-provider cross-context target check. When tools.message.crossContext.allowWithinProvider is disabled, an admitted...

CVE-2026-100553

OpenClaw versions >= 2026.6.9 and

EUVD-2026-87497

OpenClaw versions = 2026.6.9 and 2026.8.1 do not declare the native chatId parameter as a delivery target in the Feishu unpin feature, so unpin requests can bypass the shared same-provider cross-context target check. When tools.message.crossContext.allowWithinProvider is disabled, an admitted...

CVE-2026-100553 OpenClaw 2026.6.9 before 2026.8.1 Cross-Context Policy Bypass via Feishu unpin

OpenClaw versions = 2026.6.9 and 2026.8.1 do not declare the native chatId parameter as a delivery target in the Feishu unpin feature, so unpin requests can bypass the shared same-provider cross-context target check. When tools.message.crossContext.allowWithinProvider is disabled, an admitted...

EUVD-2026-87520

OpenClaw versions before 2026.8.1 fail to bind working directory context to reusable exec approvals, allowing approved commands to execute in different directories. Attackers with an allow-always approval can reuse it to run the same command against unreviewed files or repositories with materiall...

CVE-2026-100530

OpenClaw versions before 2026.8.1 fail to bind working directory context to reusable exec approvals, allowing approved commands to execute in different directories. Attackers with an allow-always approval can reuse it to run the same command against unreviewed files or repositories with materiall...

CVE-2026-100530

Versions of OpenClaw before 2026.8.1 are vulnerable to an issue where the software fails to bind the working directory context to reusable exec approvals . This oversight allows approved commands to execute in different directories than originally intended. An attacker possessing an allow-always ...

CVE-2026-100530 OpenClaw before 2026.8.1 Exec Approval Directory Binding

OpenClaw versions before 2026.8.1 fail to bind working directory context to reusable exec approvals, allowing approved commands to execute in different directories. Attackers with an allow-always approval can reuse it to run the same command against unreviewed files or repositories with materiall...

PT-2026-99190

Name of the Vulnerable Software and Affected Versions OpenClaw versions 2026.6.9 through 2026.8.0 Description The Feishu unpin feature fails to declare the native chatId parameter as a delivery target. This allows unpin requests to bypass the shared same-provider cross-context target check. When...