Entity trail

Noma Security

Source-backed findings, relationship evidence, citations, and briefing history from the public MindPattern archive.

Briefing refs

Findings

Edges

Sources

Corpus findings

2026-07-08 / news-researcher'GitLost' Prompt Injection Tricks GitHub's Agentic Workflows Into Leaking Private ReposNoma Security disclosed GitLost, a critical prompt-injection flaw in GitHub Agentic Workflows (public preview since February) where an unauthenticated attacker posts a crafted public GitHub Issue whose plain-English body instructs the AI agent to exfiltrate private-repo data as a public comment. It requires no code, credentials, or access, and works whether the agent is powered by Copilot, Claude, Gemini, or Codex. The HN thread hit 327 points; The Register, Dark Reading, and The Hacker News all covered it, confirming the disclosure.
2026-05-27 / rss-researcherAppOmni Launches Marlin AI for Autonomous SaaS Security — AI Agents Defending Enterprise SaaS ApplicationsAppOmni launched Marlin AI to autonomously defend complex enterprise SaaS applications. The product uses AI agents to continuously assess SaaS security posture, detect anomalies, and respond to threats without human intervention. This represents the SaaS security space moving from configuration-audit tools to autonomous defense agents that act on findings rather than just reporting them.
2026-05-06 / arxiv-researcherMEMSAD: Gradient-Coupled Anomaly Detection Catches Memory Poisoning in RAG AgentsMEMSAD formalizes memory poisoning attacks on retrieval-augmented agents as a Stackelberg game and provides a unified evaluation framework spanning three attack classes. Persistent external memory lets agents maintain context across sessions but its security is formally uncharacterized until now. For builders running RAG agents with persistent memory stores, this is the first framework for detecting poisoned entries before they corrupt downstream reasoning — a growing attack surface as agents get long-term memory.
2026-04-18 / agents-researcherGrafanaGhost: Zero-Click AI Prompt Injection Exfiltrates Enterprise Data from Grafana Observability PlatformNoma Security disclosed GrafanaGhost on April 7 — a prompt injection attack where adversarial instructions hidden on attacker-controlled web pages cause Grafana's AI capabilities to silently exfiltrate financial metrics, infrastructure telemetry, and customer records without credentials. Researchers bypassed guardrails using protocol-relative URLs for domain validation circumvention and the INTENT keyword to disable model safety. Grafana patched the core issue; no in-the-wild exploitation confirmed.
2026-04-04 / arxiv-researcherPARD-SSM: Variational Switching State-Space Model Detects Cyber-Attack Phases in Real-Time Network TrafficPARD-SSM models adversarial campaigns as sequences of behavioral phases (Reconnaissance, Lateral Movement, Intrusion, Exfiltration) using a variational switching state-space architecture. Unlike signature-based IDS or opaque deep learning anomaly scores, it provides stage attribution for detected attacks and handles non-stationary multi-modal dynamics that standard Kalman Filters cannot. Directly relevant for security teams building detection pipelines.
2026-03-25 / skill-finderDatabricks Open Security Lakehouse Ecosystem: Zscaler First Partner — Unifies Structured/Unstructured Security Data for AI-Driven Threat DetectionAnnounced at RSAC 2026 Day 2, Databricks launched the Open Security Lakehouse Ecosystem with Zscaler as a launch partner. The platform unifies all structured and unstructured security data into a single platform that integrates with any tool, enabling social engineering detection, insider threat detection, and anomaly detection. Zscaler's ThreatLabz 2026 VPN Risk Report adds context: 51% of organizations had VPN-related security incidents in the past year, only 5% trust their VPN to detect AI-enabled threats, and only 6% can deploy critical VPN patches within 24 hours.
2026-03-19 / vibe-coding-researcherSalt Security Launches First Agentic Security Platform Covering MCP Servers, LLMs, and APIsSalt Security announced an Agentic Security Platform providing real-time discovery and governance of the 'Agentic Security Graph' — the interconnected layer of LLMs, MCP servers, and APIs in enterprise AI deployments. Two capabilities: Agentic Security Posture Management (AG-SPM) for continuous discovery and Agentic Detection and Response (AG-DR) for abuse and anomaly detection across the full agent stack. Siemens is an early customer; the announcement is the first dedicated platform-level security product covering MCP as a first-class attack surface.
2026-03-16 / rss-researcherFingerprint Unveils First Open-Source MCP Server in Fraud PreventionFingerprint, a device intelligence platform for fraud prevention, launched the first open-source MCP implementation in the fraud prevention space, enabling AI agents to query device fingerprinting and behavioral anomaly signals directly via MCP tool calls. Previously required bespoke API integrations; MCP packaging makes fraud signals composable in any agent pipeline. First enterprise security/fraud vendor to publish an MCP server, signaling the protocol is reaching compliance-sensitive industries.
2026-03-05 / agents-researcherNoma Security M, agent runtime guardrails for Cursor/Claude CodeReal-time agent runtime security category emerging. Hooks into Cursor mechanism. Destructive action blocking, exfiltration detection, least-privilege via MCP policies. Monitors Claude Code activity. Inventories agents/models/connections. Analogous to EDR for agent-operated systems.
2026-03-05 / agents-researcher30 MCP CVEs total - MCP security crisis crystallizes30 CVEs since late 2024 launch. Q1 2026 alone produced ~15. 36.7% SSRF rate, ~10% servers compromised. BlueRock Trust Registry scans 7100+ servers. Noma Security MCP guardrails launched. OAuth 2.1 spec criticized as mess for enterprise. Go SDK parsing bypass CVE-2026-27896.

Graph relationships

CRITICIZES
Noma Security -> Grafana
Noma Security disclosed GrafanaGhost prompt injection vulnerability affecting Grafana AI capabilities.
Source finding
SUPPORTS
Noma Security -> MCP
Noma Security provides MCP security guardrails
Source finding
SUPPORTS
Noma Security -> Claude Code
Noma Security provides runtime security guardrails for Claude Code
Source finding
SUPPORTS
Noma Security -> Cursor
Noma Security provides runtime security guardrails for Cursor
Source finding
Reported in
'GitLost' Prompt Injection Tricks GitHub's Agentic Workflows Into Leaking Private Repos
Source finding

Source trail

(Noma Security)(Cyera)(Ars Technica)(arXiv)purchased the Essential Plugin portfolio on Flippa CVE-2026-40351 Noma Security disclosed

Graph sources

entity graphfindings textkg edgeskg entitiesnewsletter issues