These pages are published after PatchSiren validates generated defensive summaries against stored public CVE and source evidence.
AI-assisted PatchSiren debrief based on the supplied source corpus. The CVE record was published on 2026-08-13T12:17:26.337Z and has not been modified since then. Network-AI versions before 5.15.1 contain a security matcher bypass vulnerability where SandboxPolicy evaluates raw command strings with quotes preserved while the executor tokenizes commands by stripping quotes before execution. This discrepanc [truncated]
Network-AI is a TypeScript/Node.js multi-agent orchestrator that is vulnerable to an arbitrary path deletion issue. The `EnvironmentManager.listBackups()` function reads each backup's `_manifest.json` and trusts the manifest's `path` field. This could allow an attacker to cause `EnvironmentManager.pruneBackups()` to recursively delete an arbitrary path accessible to the Network-AI process user. The issue [truncated]
Network-AI's multi-agent orchestrator features a shipped, exported, documented `ApprovalInbox` (`lib/approval-inbox.ts`) with a critical vulnerability (CVE-2026-58482). This component, part of the human-in-the-loop Approval Gate, exposes an HTTP server without authentication and with a wildcard CORS configuration (`Access-Control-Allow-Origin: *`). Consequently, in versions 5.0.0 through 5.12.1, any entit [truncated]
Network-AI is a TypeScript/Node.js multi-agent orchestrator that allows for sandbox escape due to improper path containment checks in its AgentRuntime. An attacker with access to AgentRuntime's readFile() or listDir() functions can read or list files outside the intended sandbox directory. The issue is fixed in v5.12.2, where SandboxPolicy.resolvePath() and isPathAllowed() now use separator-anchored prefi [truncated]
AI-assisted PatchSiren debrief based on the supplied source corpus. The CVE record was published on 2026-07-20T17:18:15.570Z and has not been modified since then. The NVD entry is currently Analyzed. The EnvironmentManager in Network-AI is vulnerable to a directory traversal attack due to improper validation of backup IDs. This allows an attacker to restore files from an arbitrary directory into the targe [truncated]
AI-assisted PatchSiren debrief based on the supplied source corpus. The CVE record was published on 2026-07-20T17:17:09.990Z and has not been modified since then. The NVD entry is currently Analyzed. Network-AI is a TypeScript/Node.js multi-agent orchestrator. Prior to version 5.4.5, the MCP SSE server defaults to an empty secret, allowing unauthorized access and enabling an attacker to invoke exposed MCP [truncated]
CVE-2026-64623 is an improper cryptographic signature verification vulnerability in Network-AI before 5.13.4. The APSAdapter local verifier accepts any non-empty string as valid, allowing unauthenticated attackers to submit forged APS delegation payloads and obtain signed permission-grant tokens for sensitive resources like SHELL_EXEC. This vulnerability has a high CVSS score of 8.8, indicating a high sev [truncated]
The Network-AI npm package, specifically versions 5.12.2 through 5.13.3, contains a critical vulnerability. This vulnerability fails to apply the configured authorization check to certain GET read routes, allowing unauthenticated actors to access sensitive approval request details. The affected routes include GET /approvals/?status=all, GET /approvals/:id, GET /approvals/stats, and GET /approvals/sse. The [truncated]
CVE-2026-48814 is a critical vulnerability in Network-AI, a TypeScript/Node.js multi-agent orchestrator. Versions 5.7.1 and earlier are affected by an issue allowing unauthenticated cross-origin MCP tool invocation. This is due to an empty default secret used by the MCP SSE server. Although CVE-2026-46701 partially addressed this issue in version 5.4.5 by restricting CORS to localhost origins, the empty d [truncated]
