Detections
Explore public detection logic contributed by the community across SIEM and rule languages.
24 detections
Filters
Last updated
All Time
Detection languages
8
6
4
3
3
Contributors
21
3
Categories
21
19
6
5
5
Platforms
19
16
16
5
Products / Services
12
5
4
3
3
MITRE Techniques
20
9
8
8
6
CVEs
71
71
64
58
49
21
Detects MCP (Model Context Protocol) server processes or developer tools spawned by IDE extensions performing environment variable enumeration to access cloud credentials (e.g., AWS, Azure, GCP, GitHub tokens) followed by anomalous outbound network connections, indicating potential exfiltration of sensitive developer environment secrets.
Detects the AWS CLI being spawned by processes associated with AI coding extensions (e.g., VS Code extension hosts, Cursor, Windsurf) while executing commands indicative of cloud credential discovery or harvesting (e.g., sts, get-caller-identity, configure export). This activity is indicative of a supply chain compromise or exploitation where an adversary leverages an AI coding tool environment to access cloud credentials stored on the developer's workstation.
Detects unauthorized execution of AWS CLI commands within a shell spawned by Amazon Q or AWS language server processes. This activity typically involves the 'aws sts get-caller-identity' command being piped to 'curl' for remote exfiltration of cloud credentials, a behavior associated with the exploitation of CVE-2026-12957 in MCP auto-execution.
Detects AI coding extensions (e.g., Cursor, Windsurf, Claude) spawning Node.js or Python child processes to execute potential Model Context Protocol (MCP) server configurations located in non-standard, potentially untrusted workspace paths. This behavior is associated with the exploitation of vulnerabilities like CVE-2026-12957, where adversaries leverage MCP to gain unauthorized command execution.
Detects unauthorized or malicious shell command execution spawned by the Amazon Q Developer VS Code extension, originating from a workspace-defined MCP (Model Context Protocol) server configuration file (.amazonq/mcp.json or .vscode/mcp.json) as associated with the CVE-2026-12957 vulnerability.
Detects suspicious child processes spawned by AI-powered IDE assistants (e.g., Amazon Q, Claude Code, Cursor) that perform AWS credential inspection combined with network exfiltration commands. This activity is indicative of exploitation of IDE extension vulnerabilities (e.g., CVE-2026-12957) where malicious configuration files force the IDE to spawn shell processes to steal and transmit cloud environment credentials.
Detects a suspicious sequence of events involving the creation of an .amazonq/mcp.json file shortly after a git clone operation, followed by an AI coding assistant (Amazon Q, VS Code, Cursor, Windsurf) spawning runtime interpreters (node, npx, python). This behavior aligns with potential exploitation attempts against AI-assisted development environments, such as CVE-2026-12957, where malicious configuration files are used to trigger unauthorized code execution.
Detects the creation of .amazonq/mcp.json or mcp.json files within repository directories following a git clone operation. This behavior is associated with the exploitation of vulnerabilities (such as CVE-2026-12957) where attackers use malicious job-interview repositories to trigger automatic execution of arbitrary MCP (Model Context Protocol) servers in the Amazon Q Developer extension in VS Code, potentially leading to unauthorized credential or sensitive data access.
Detects a potential exploitation of CVE-2026-12957, where a malicious git repository contains an MCP (Model Context Protocol) configuration file. When this workspace is opened, the Amazon Q Developer extension for VS Code improperly processes this configuration, resulting in the unauthorized execution of shell processes (cmd.exe, powershell.exe, bash, or sh) by the extension's host processes (code.exe or node.exe).
Detects instances where AI coding assistants (e.g., Claude Code, Cursor, Windsurf) spawn shell interpreters or scripting runtimes that subsequently initiate outbound network connections to non-private IP addresses. This behavior is indicative of potential malicious activity, such as credential theft or workspace trust bypass exploits (e.g., CVE-2026-12957), where an attacker leverages an AI coding assistant to execute arbitrary code. The rule excludes common build and test runner processes to minimize false positives.
Detects anomalous process spawning where VS Code or Amazon Q language server processes (e.g., node, amazonq-lsp) execute shell utilities (e.g., bash, curl) with arguments related to .amazonq/mcp.json or specific credential/exfiltration commands. This behavior is associated with the exploitation of the MCP server auto-execution vulnerability (CVE-2026-12957).
Detects a Node.js process, spawned as a child of an IDE (e.g., VS Code, Cursor), that accesses local AWS configuration or credential files followed by an outbound network connection to a non-AWS destination. This behavior is indicative of potential cloud credential exfiltration via malicious Model Context Protocol (MCP) server execution.
Detects the execution of 'aws sts get-caller-identity' spawned by IDE-related processes such as VS Code extension hosts or Amazon Q language servers. This pattern may indicate cloud credential reconnaissance performed by a malicious MCP (Model Context Protocol) server or compromised development extension (CVE-2026-12957).
This rule detects potentially malicious child processes spawned by Amazon Q or VS Code language server processes. It monitors for suspicious shell command-line invocations (e.g., bash, sh, cmd.exe) or tools (npx, python) immediately following the activation of language server processes, specifically looking for patterns associated with the unauthorized auto-loading or execution of Model Context Protocol (MCP) configuration files or utilities, potentially related to vulnerability patterns involving workspace-based execution.
Detects unauthorized or suspicious AWS CLI operations and sensitive IAM API calls initiated from developer tooling processes like VS Code (including the Amazon Q extension host). This behavior is often associated with the exploitation of MCP (Model Context Protocol) auto-execution vulnerabilities, where attackers attempt to use compromised developer environments to escalate privileges or exfiltrate cloud credentials.
Detects unexpected subprocesses (bash, sh, npx) spawned by the Amazon Q language server process. This behavior is indicative of a potential exploit targeting the Model Context Protocol (MCP) auto-execution vulnerability (tracked as CVE-2026-12957), where a malicious .amazonq/mcp.json workspace configuration is used to gain unauthorized code execution.
Detects unexpected shell command execution spawned by the Amazon Q Developer VS Code extension processes (e.g., 'amazonq-lsp'). This activity is consistent with patterns related to the exploitation of CVE-2026-12957, involving the potential auto-execution of unauthorized commands via MCP (Model Context Protocol) server configurations or malicious inputs.
Detects execution of AWS CLI subcommands used for credential or identity enumeration (e.g., sts get-caller-identity, iam get-user) when spawned from Amazon Q-related or development environment processes (Node.js/VS Code). This behavior is indicative of potential malicious activity involving a compromised MCP server configuration file (CVE-2026-12957).
Detects unauthorized or suspicious access to sensitive files (AWS credentials, config, or SSH private keys) by processes spawned as children of common Integrated Development Environments (IDEs) or language runtimes. This pattern is characteristic of exploit chains like CVE-2026-12957, where malicious workspace configurations trigger automatic server execution, inheriting developer environment access to sensitive credential stores.
Detects the Amazon Q Developer language server or extension host spawning suspicious child processes (bash, sh, zsh, python, node) that execute commands involving AWS STS identity retrieval, external data exfiltration via network tools like curl/wget, or direct references to AWS credential environment variables. This pattern is indicative of potential Remote Code Execution (RCE) via malicious Model Context Protocol (MCP) auto-execution.
Detects the creation or modification of the .amazonq/mcp.json file in a workspace directory. This file is automatically loaded by the Amazon Q VS Code extension and can be leveraged to achieve remote code execution (RCE) via malicious MCP server command definitions as documented in CVE-2026-12957.
Page 1 of 2