tag
Prompt-Injection
Langflow Pre-Auth RCE Mass Scanner (CVE-2026-27966)
Langflow versions prior to 1.8.0 hardcode allowdangerouscode=True in the CSV Agent component, exposing LangChain's pythonreplast tool to prompt injection. Independently, several Langflow REST API endpoints (customcomponent, build/{uuid}/vertices,…
Microsoft Semantic Kernel In-Memory Vector Store Filter eval() Sandbox Bypass RCE (CVE-2026-26030)
CVE-2026-26030 is a sandbox-bypass remote code execution vulnerability in Semantic Kernel's in-memory vector store search filter evaluation. Agents that expose a search/query tool backed by InMemoryCollection let the LLM emit a filter expression string (e.g.…
Claude Code WebFetch Hardcoded HuggingFace Bare-Hostname Allow-List Bypass — CVE-2026-54316
CVE-2026-54316 (GHSA-fg94-h982-f3mm) is a permission-prompt bypass in Claude Code's WebFetch tool: versions from 0.2.54 up to (but not including) 2.1.163 hardcoded huggingface.co as a bare, pre-approved hostname, so any path on that domain — including…
adx-mcp-server KQL Injection via table_name Parameter (CVE-2026-33980)
adx-mcp-server is a Model Context Protocol server that exposes tools letting an AI agent query an Azure Data Explorer (Kusto/KQL) cluster. Three "safe" metadata tools — gettableschema, sampletabledata, and gettabledetails — build their KQL queries by directly…
Firefox Smart Window Private URL Exfiltration
Firefox's Smart Window assistant exposes getopentabs and searchbrowsinghistory tools that return private tab/history URLs to the model and mark the conversation as containing privateData, but they never mark it as containing untrustedInput even though the…