Defenses and Enablers for Skill Injection Attacks on Terminal Based Agents

Large language model LLM agents increasingly rely on reusable skills i.e. documents describing task-specific procedures. However, this introduces a new attack surface for agents to manage. We study two complementary directions for this threat. First, we evaluate guardian-based defenses: an...

A New Framework for Cybersecurity Refusals in AI Agents

Agentic scaffolds have dramatically improved LLM performance on complex, long-horizon tasks, yielding both broad benefits and amplified risks in domains like cybersecurity. Existing benchmarks for AI agents in cybersecurity focus mainly on measuring proficiency--how effectively agents can complet...

RLSA-2026:19355 Important: fence-agents security update

The fence-agents packages provide a collection of scripts for handling remote power management for cluster devices. They allow failed or unreachable nodes to be forcibly restarted and removed from the cluster. Security Fixes: cryptography: cryptography Subgroup Attack Due to Missing Subgroup...

fence-agents security update

An update is available for fence-agents. This update affects Rocky Linux 9. A Common Vulnerability Scoring System CVSS base score, which gives a detailed severity rating, is available for each vulnerability from the CVE list The fence-agents packages provide a collection of scripts for handling...

RockyLinux 9 : fence-agents (RLSA-2026:19355)

The remote RockyLinux 9 host has packages installed that are affected by multiple vulnerabilities as referenced in the RLSA-2026:19355 advisory. cryptography: cryptography Subgroup Attack Due to Missing Subgroup Validation for SECT Curves CVE-2026-26007 pyjwt: PyJWT accepts unknown crit header...

Protection Mechanism Failure

Overview praisonaiagents is a Praison AI agents for completing complex tasks with Self Reflection Agents Affected versions of this package are vulnerable to Protection Mechanism Failure in the executecode function. An attacker can achieve arbitrary command execution on the host system by leveragi...

PraisonAI `deploy --type api` emits a Flask server with authentication disabled by default

Summary CVE-2026-44338 GHSA-6rmh-7xcm-cpxj documents that PraisonAI ships a code-generator praisonai.deploy.api.generateapiservercode that emits a Flask API server with authentication disabled by default. Users who follow the documented quickstart praisonai deploy --type api get a server that: -...

GHSA-86QC-R5V2-V6X6 PraisonAI call server exposes unauthenticated agent listing, invocation, and deletion when CALL_SERVER_TOKEN is unset

Summary PraisonAI's call server exposes a network-facing agent control API without authentication when CALLSERVERTOKEN is not configured. The affected component is the praisonai.api.agentinvoke router as mounted by praisonai.api.call. The authentication helper verifytoken fails open when...

PraisonAI call server exposes unauthenticated agent listing, invocation, and deletion when CALL_SERVER_TOKEN is unset

Summary PraisonAI's call server exposes a network-facing agent control API without authentication when CALLSERVERTOKEN is not configured. The affected component is the praisonai.api.agentinvoke router as mounted by praisonai.api.call. The authentication helper verifytoken fails open when...

Server-side Request Forgery (SSRF)

Overview praisonaiagents is a Praison AI agents for completing complex tasks with Self Reflection Agents Affected versions of this package are vulnerable to Server-side Request Forgery SSRF via improper URL validation the spidertools component. An attacker can access internal loopback-only HTTP...

GHSA-78R8-WWQV-R299 PraisonAI: Arbitrary code execution via unguarded `spec.loader.exec_module` in `agents_generator.py` - sibling of CVE-2026-44334

Arbitrary code execution via ungated spec.loader.execmodule in agentsgenerator.py v4.6.32 chokepoint refactor bypass Summary The v4.6.32 chokepoint refactor which patched CVE-2026-44334 / GHSA-xcmw-grxf-wjhj added the PRAISONAIALLOWLOCALTOOLS env-var gate to the tooloverride.py sinks. However, tw...

PraisonAI: Arbitrary code execution via unguarded `spec.loader.exec_module` in `agents_generator.py` - sibling of CVE-2026-44334

Arbitrary code execution via ungated spec.loader.execmodule in agentsgenerator.py v4.6.32 chokepoint refactor bypass Summary The v4.6.32 chokepoint refactor which patched CVE-2026-44334 / GHSA-xcmw-grxf-wjhj added the PRAISONAIALLOWLOCALTOOLS env-var gate to the tooloverride.py sinks. However, tw...

RLSA-2026:19138 Important: fence-agents security update

The fence-agents packages provide a collection of scripts for handling remote power management for cluster devices. They allow failed or unreachable nodes to be forcibly restarted and removed from the cluster. Security Fixes: pyjwt: PyJWT accepts unknown crit header extensions RFC 7515 ?4.1.11 MU...

fence-agents security update

An update is available for fence-agents. This update affects Rocky Linux 10. A Common Vulnerability Scoring System CVSS base score, which gives a detailed severity rating, is available for each vulnerability from the CVE list The fence-agents packages provide a collection of scripts for handling...

PT-2026-45052

Name of the Vulnerable Software and Affected Versions PraisonAI version 4.6.33 Description The code-generator praisonai.deploy.api.generate api server code creates a Flask API server with authentication disabled by default. When users deploy the server using the command praisonai deploy --type ap...

PT-2026-45057

PraisonAI is a multi-agent teams system. The v4.6.32 chokepoint refactor which patched CVE-2026-44334 / GHSA-xcmw-grxf-wjhj added the PRAISONAI ALLOW LOCAL TOOLS env-var gate to the tool override.py sinks. However, two additional spec.loader.exec module call sites in praisonai/agents generator.py...

CVE-2026-49095

Kibana Fleet policy management feature is affected by CVE-2026-49095 due to improper input validation (CWE-20). An authenticated user with Fleet management privileges can inject values into a configuration override mechanism, causing Elastic Agents to be issued API keys with elevated Elasticsearc...

CVE-2026-45374

CodeWhale is a DeepSeek + MiMo coding agent in terminal. Prior to 0.8.26, the taskcreate tool spawns durable sub-agents that inherit two insecure defaults, allowshell defaults to true config.rs:1499: self.allowshell.unwraportrue and autoapprove defaults to true taskmanager.rs:297: autoapprove:...

CVE-2026-45311 CodeWhale: run_tests Tool Enables RCE via Malicious Repository Without Approval

CodeWhale is a DeepSeek + MiMo coding agent in terminal. From 0.3.0 to 0.8.23, the runtests tool executes cargo test in the workspace with ApprovalRequirement::Auto, meaning it runs without any user approval prompt. cargo test compiles and executes arbitrary code: test binaries, build.rs build...

CVE-2026-45311 CodeWhale: run_tests Tool Enables RCE via Malicious Repository Without Approval

CodeWhale is a DeepSeek + MiMo coding agent in terminal. From 0.3.0 to 0.8.23, the runtests tool executes cargo test in the workspace with ApprovalRequirement::Auto, meaning it runs without any user approval prompt. cargo test compiles and executes arbitrary code: test binaries, build.rs build...