A weakness has been identified in huggingface smolagents 1.25.0.dev0. This affects the function evaluate_augassign/evaluate_call/evaluate_with of the file src/smolagents/local_python_executor.py of the component Incomplete Fix CVE-2025-9959. This manipulation causes code injection. It is possible to initiate the attack remotely. The exploit has been made available to the public and could be used for attacks. The vendor was contacted early about this disclosure but did not respond in any way.

A weakness has been identified in huggingface smolagents 1.24.0. Impacted is the function requests.get/requests.post of the component LocalPythonExecutor. Executing a manipulation can lead to server-side request forgery. It is possible to launch the attack remotely. The exploit has been made available to the public and could be used for attacks. The vendor was contacted early about this disclosure but did not respond in any way.

Summary A Reflected Cross-Site Scripting (XSS) vulnerability was discovered in the AI Playground's OAuth callback handler. The `error_description` query parameter was directly interpolated into an HTML script tag without proper escaping, allowing attackers to execute arbitrary JavaScript in the context of the victim's session. Root cause The OAuth callback handler in `site/ai-playground/src/server.ts` directly interpolated the `authError` value, sourced from the `error_description` query parameter, into an inline `<script>` tag. Impact An attacker could craft a malicious link that, when clicked by a victim, would: * Steal user chat message history - Access all LLM interactions stored in the user's session. * Access connected MCP Servers - Interact with any MCP servers connected to the victim's session (public or authenticated/private), potentially allowing the attacker to perform actions on the victim's behalf Mitigation: * PR:  https://github.com/cloudflare/agents/pull/841 https://github.com/cloudflare/agents/pull/841 * Agents-sdk users should upgrade to agents@0.3.10 * Developers using configureOAuthCallback with custom error handling in their own applications should ensure all user-controlled input is escaped before interpolation.

Summary An Insecure Direct Object Reference has been found to exist in `createHeaderBasedEmailResolver()` function within the Cloudflare Agents SDK. The issue occurs because the `Message-ID` and `References` headers are parsed to derive the target agentName and agentId without proper validation or origin checks, allowing an external attacker with control of these headers to route inbound mail to arbitrary Durable Object instances and namespaces . Root cause The `createHeaderBasedEmailResolver()` function lacks cryptographic verification or origin validation for the headers used in the routing logic, effectively allowing external input to dictate internal object routing. Impact Insecure Direct Object Reference (IDOR) in email routing lets an attacker steer inbound mail to arbitrary Agent instances via spoofed Message-ID. Mitigation: * PR: https://github.com/cloudflare/agents/blob/main/docs/email.md ] provides the necessary architectural context for coding agents to mitigate the issue by refactoring the resolver to enforce strict identity boundaries. * Agents-sdk users should upgrade to agents@0.3.7

Incomplete validation of dunder attributes allows an attacker to escape from the Local Python execution environment sandbox, enforced by smolagents. The attack requires a Prompt Injection in order to trick the agent to create malicious code.