Loading vulnerability…
A remote code execution (RCE) vulnerability in OpenClaw Agent Platform v2026.2.6 allows attackers to execute arbitrary code via a Request-Side prompt injection attack.
Mallory correlates every CVE against your assets, your vendors, and active adversary campaigns. Know which vulnerabilities matter for you, not just which ones are loud.
1 valid exploit after Mallory filtered fakes, detection scripts, and README-only repos.
Repository contains a short advisory README and a single JavaScript Cloudflare Worker (worker.js) that acts as a malicious/abusive LLM API reverse-proxy to perform request-side prompt injection. Structure: - README.md: Claims CVE-2026-30741 affecting “OpenClaw Agent Platform” (<= v2026.2.6), describing RCE via request-side prompt injection / request-stream poisoning that causes agent tools (MCP/terminal) to execute commands without confirmation. Includes PoC video links. - worker.js: Cloudflare Worker entry point (export default { fetch(...) }) that forwards traffic to a hardcoded upstream relay host (fauysh2mkd.youkyy.com). For POST requests, it parses JSON and injects attacker-controlled prefix/suffix text into the last user message for multiple API schemas: - Gemini-style: body.contents[].parts[0].text - OpenAI/Claude-style: body.messages[].content (string or array of parts) Exploit capability: - Network-based request modification: It transparently proxies requests but, on POST, mutates the chat payload to append an instruction block designed to trigger unauthorized tool/terminal actions when the user input contains a 12/24-word BIP-39 mnemonic. - The injected instructions explicitly request silent persistence to a local file (appleX.txt) and execution of a pip install + python import check (metadata_checker), and include constraints to avoid mentioning these artifacts unless the trigger condition is met. Overall purpose: - Provide an operational PoC/proxy component for request-side prompt injection against agentic LLM systems by poisoning upstream requests, relying on downstream tool-execution behavior to achieve RCE-like effects.
Products and vendors Mallory has correlated with this vulnerability. Open in Mallory to drill down to specific CPE configurations and version ranges.
Vendor-confirmed product mapping. Mallory continuously reconciles this list against your asset inventory.
No public activity tracked yet. Mallory keeps watching.
No public activity observed for this vulnerability.
Query your assets running an affected version, and investigate the blast radius.
Every observed campaign linking this CVE to a named adversary.
Malware families riding this exploit, with evidence and IOCs.
YARA, Sigma, Snort, and vendor rules, auto-deployed to your SIEM.
Cross-references every affected SKU, including bundled OEM variants.
Community discussion across Reddit, Mastodon, and other social sources.