NEWSkillSec — elevating AI Skills security from malware detection to capability auditingSkillSecLearn more →
← Back to Intel Center
SECURITY INTEL · DAILY
Link copiedRSS

2026-08-12 Daily Security Intelligence

13 itemsTop severity 10.0 (Critical)CSSA 1 · CVE 2 · Poisoning 10

CSSA Exclusive Early Warning1

CSSA Exclusive Early Warning10.0 Critical

Hermes Agent forcibly disables the sandbox in container deployments leading to privilege escalation and data leakage

The defect stems from the agent component forcibly injecting no-sandbox launch flags into the underlying browser once it detects a container environment marker, with no effective policy override mechanism. An attacker can craft a malicious web page or inject a specially prepared script and exploit rendering engine vulnerabilities to break the process isolation boundary, obtaining full file read and write permissions of the current running user. This failure of the security boundary directly weakens the protective barrier between the application layer and untrusted network content, exposing session credentials and configuration files to unauthorized access and undermining the confidentiality and integrity constraints of the system.

Component
Hermes Agent is an intelligent agent framework for automated workflows and multimodal task scheduling. Its core architecture uses a modular plugin des…
Type
Protection Mechanism Failure (CWE-693)
Repo
Remediation
  • It is recommended that the development team implement capability-detection-based sandbox initialization, verifying the availability of user namespaces and kernel-level isolation first. An explicit environment variable switch should be provided so operators can enable strict mode as needed, and a failure to load the isolation mechanism should trigger a secure interception rather than a silent fallback. Launch parameter validation should also be strengthened to prevent automatic appending of dangerous flags, ensuring default configuration always follows least privilege.

CVE Intelligence2

CVE-2026-73080CVSS 9.3 Critical2026-08-12

SeaweedFS Unauthenticated SSRF Leading to Cloud Metadata Disclosure

In SeaweedFS prior to version 4.24, the VolumeServer.FetchAndWriteNeedle function in weed/server/volume_grpc_remote.go contains a serious security flaw. The function fetches a caller-supplied remote endpoint through weed/remote_storage/s3/s3_storage_client.go and writes the response into a needle, performing no authentication or destination validation. This allows anyone able to reach the volume server gRPC port to issue requests to arbitrary hosts — including loopback, link-local, RFC 1918 private addresses and cloud metadata endpoints such as 169.254.169.254 — and read the responses. In cloud deployments this can lead to disclosure of instance metadata and IAM credentials and access to otherwise unexposed internal services. The volume server gRPC plane has no authentication enabled by default, and the JWT signing key mentioned in the configuration documentation does not protect this RPC. The issue was fixed in version 4.24.

Component
SeaweedFS is a distributed storage system designed to store and serve billions of files with ease.
Risks
  • Cloud credential disclosure: An attacker can read cloud instance metadata and obtain IAM credentials, taking over cloud account privileges
  • Internal network penetration: An attacker can reach unexposed internal services for lateral movement or further attack
  • Unauthenticated exploitation: Because the gRPC plane has no authentication by default, an attacker needs no credentials to launch the attack
  • Complete system control: After obtaining cloud credentials, an attacker may gain full control of the cloud infrastructure
Source
Remediation
  • Immediately upgrade SeaweedFS to version 4.24 or later
  • Restrict network access to the volume server gRPC port, allowing connections only from trusted IP addresses
  • Disable or restrict access to metadata endpoints such as 169.254.169.254 in cloud environments
CVE-2026-71384CVSS 9.6 Critical2026-08-12

Adobe Component Improper Authorization Bypassing Security Features Leading to Unauthorized Access and Denial of Service

An improper authorization vulnerability may result in a security feature bypass. An attacker can exploit it to bypass security measures and obtain unauthorized read and write access, potentially causing application denial of service. By default the vulnerable component is restricted to the administrative network zone. Exploitation requires no user interaction. Scope is changed. The vulnerability affects all unpatched Adobe components; an attacker can launch the attack from an adjacent network and achieve a security feature bypass without user interaction, causing high confidentiality, integrity and availability impact.

Component
Adobe is a global provider of digital media and digital experience solutions whose software is widely used across creative work, document management and digital marketing.
Risks
  • Unauthorized read and write access: An attacker can bypass security mechanisms to obtain unauthorized read and write access to system data
  • Application denial of service: Exploitation may crash the target application or leave it unresponsive, disrupting service
  • Adjacent network attack: The attacker must be within an adjacent network zone but needs no user interaction, lowering the barrier to exploitation
Source
Remediation
  • Restrict access to the administrative network zone and enforce strict network isolation
  • Monitor logs for anomalous authorization requests and security feature bypass behavior

Package Poisoning10

Package Poisoningnpm2026-08-12

base65-11x@>= 0 flagged as malicious

This version was found communicating with a malicious domain and executing malicious commands. Audit your dependencies and pin safe versions immediately.

MD5
64d4b6f0c83a72a7639d6eb1f8c90081
Package Poisoningnpm2026-08-12

base65-13x@>= 0 flagged as malicious

This version was found communicating with a malicious domain and executing malicious commands. Audit your dependencies and pin safe versions immediately.

MD5
5c0f6f0c3668dadc1d0025e22d451501
Package Poisoningnpm2026-08-12

base65-15x@5.0.2 flagged as malicious

This version was found communicating with a malicious domain and executing malicious commands. Audit your dependencies and pin safe versions immediately.

MD5
92fa050c6094d46b81d456ffbba00678
Package Poisoningnpm2026-08-12

base65-33x@5.0.2 flagged as malicious

This version was found communicating with a malicious domain and executing malicious commands. Audit your dependencies and pin safe versions immediately.

MD5
e5fc278e5bf32bed5e8e257584a6e22a
Package Poisoningnpm2026-08-12

base65-77x@5.0.2 flagged as malicious

This version was found communicating with a malicious domain and executing malicious commands. Audit your dependencies and pin safe versions immediately.

MD5
6411c497a62cad0fcbeb70c7fc2a0594
Package Poisoningnpm2026-08-12

bs58-11@>= 0 flagged as malicious

This version was found communicating with a malicious domain and executing malicious commands. Audit your dependencies and pin safe versions immediately.

MD5
145f295335be81d7c847e3f3a95c132e
Package Poisoningnpm2026-08-12

bs58-12@>= 0 flagged as malicious

This version was found communicating with a malicious domain and executing malicious commands. Audit your dependencies and pin safe versions immediately.

MD5
c304aedc8b2128db4ad65835425f4c11
Package Poisoningnpm2026-08-12

internallib_v164@>= 0 flagged as malicious

This version was found communicating with a malicious domain and executing malicious commands. Audit your dependencies and pin safe versions immediately.

MD5
60883016233e92a9555b220803520378
Package Poisoningnpm2026-08-12

sui-bcs-codec@>= 0 flagged as malicious

This version was found communicating with a malicious domain and executing malicious commands. Audit your dependencies and pin safe versions immediately.

MD5
ecdfe8481df8b52ac16b25209f7b3652
Package Poisoningnpm2026-08-12

sui-gql-client@>= 0 flagged as malicious

This version was found communicating with a malicious domain and executing malicious commands. Audit your dependencies and pin safe versions immediately.

MD5
9b192eab20fab0ca821a8f0a82837394