Sebastion
←All topics

ai

41 pieces of writing

Softeria ms-365-mcp-server PR #456 validates redirect_uri before Microsoft Entra forwarding
vulnerability7 min read

Softeria ms-365-mcp-server PR #456 validates redirect_uri before Microsoft Entra forwarding

Softeria's ms-365-mcp-server forwarded client-supplied OAuth redirect_uri values to Microsoft Entra without local validation. PR #456 adds scheme checks, loopback-only HTTP defaults and an exact-match allowlist for hosted deployments.

CodeGraphContext PR #882 rejects write Cypher on /api/graph
vulnerability7 min read

CodeGraphContext PR #882 rejects write Cypher on /api/graph

mcp-searxng PR #71 fixed a CWE-1333 ReDoS in section extraction
vulnerability7 min read

mcp-searxng PR #71 fixed a CWE-1333 ReDoS in section extraction

Vercel breached through a compromised Context.ai OAuth grant
security10 min read

Vercel breached through a compromised Context.ai OAuth grant

A compromised AI productivity tool called Context.ai gave attackers OAuth access to a Vercel employee's Google Workspace, pivoting into internal systems. The AI tool supply chain is the new CI/CD supply chain.

From tj-actions to LiteLLM to MCP: supply chain compromise now operates at infrastructure scale
security9 min read

From tj-actions to LiteLLM to MCP: supply chain compromise now operates at infrastructure scale

NVIDIA's RAG Blueprint had a path traversal in its MCP server. We got the fix merged in three days.
vulnerability6 min read

NVIDIA's RAG Blueprint had a path traversal in its MCP server. We got the fix merged in three days.

Seven authentication bypasses that keep shipping in 2025 and 2026: the same architectural antipatterns, rewritten in new frameworks
security11 min read

Seven authentication bypasses that keep shipping in 2025 and 2026: the same architectural antipatterns, rewritten in new frameworks

Modern frameworks keep reimplementing the same seven authentication bypass patterns. From hardcoded credentials to missing origin checks, the bugs are structural, not accidental, and the AI tooling boom is accelerating the cycle.

Edict's file:// handler let anyone read files outside the project directory (CWE-22)
vulnerability5 min read

Edict's file:// handler let anyone read files outside the project directory (CWE-22)

AIPex's localhost daemon let any website control your browser through a WebSocket
vulnerability5 min read

AIPex's localhost daemon let any website control your browser through a WebSocket

LangFlow, n8n and the pattern where AI configuration becomes code execution
security10 min read

LangFlow, n8n and the pattern where AI configuration becomes code execution

AI orchestration platforms like LangFlow and n8n are accumulating critical RCE vulnerabilities because their architectures treat user-supplied configuration as trusted code.

Anthropic shipped its entire source code to npm and the internet kept it forever
security10 min read

Anthropic shipped its entire source code to npm and the internet kept it forever

Every MCPHub instance started with the same admin password. I changed that.
vulnerability7 min read

Every MCPHub instance started with the same admin password. I changed that.

LightRAG's Memgraph backend had a Cypher injection vulnerability hiding in plain sight
vulnerability7 min read

LightRAG's Memgraph backend had a Cypher injection vulnerability hiding in plain sight

LightRAG's Memgraph storage backend interpolated unsanitised entity types directly into Cypher queries, enabling injection via the API. The Neo4j backend was already fixed.

Environment variables are the new command line: how AI agents keep leaking secrets through configuration files
security12 min read

Environment variables are the new command line: how AI agents keep leaking secrets through configuration files

TeamPCP compromised the AI proxy that holds everyone's API keys
security9 min read

TeamPCP compromised the AI proxy that holds everyone's API keys