Network-AI: ApprovalInbox HTTP server has no authentication — anyone can approve pending agent actions
Network-AI, a TypeScript/Node.js multi-agent orchestrator, has a shipped, exported, documented feature called `ApprovalInbox` (`lib/approval-inbox.ts`). It is the network surface of the human-in-the-loop Approval Gate, which `ApprovalGate` uses to require explicit human approval for high-risk operations. The HTTP server it exposes has no authentication of any kind and sets `Access-Control-Allow-Origin: *` on every route, including the state-changing `POST /approvals/:id/approve` and `/deny`. As a result, in versions 5.0.0 through 5.12.1, any party who can send an HTTP request to the inbox port — a co-located process, a container/SSRF on the same host, a remote client when the operator binds a non-loopback address, or any website the operator visits in a browser (via the wildcard CORS) — can enumerate pending approvals and approve them, defeating the entire human-in-the-loop control and causing the gated high-risk action (e.g. a shell command the agent was holding for review) to execute without consent. This issue is fixed in v5.12.2. `ApprovalInbox` now accepts a `secret` option. When set, the mutating endpoints `POST /:id/approve` and `POST /:id/deny` require an `Authorization: Bearer ` header, validated in constant time with `crypto.timingSafeEqual`. `startServer()` already binds to `127.0.0.1` by default; operators exposing the inbox on a network must set a secret.
- Products
- Not specified
- Affected versions
- Not specified
- Fixed versions
- Not specified
- CVEs
- CVE-2026-58482
- Confirmed fixed CVEs
- Not specified
- Updated in this feed
- 8 Nov 2026, 02:41 UTC · Revision 1
Source evidence
Network-AI: ApprovalInbox HTTP server has no authentication — anyone can approve pending agent actions
core-osv · Original sourcePublished 20 Jul 2026, 16:38 UTC
Network-AI, a TypeScript/Node.js multi-agent orchestrator, has a shipped, exported, documented feature called `ApprovalInbox` (`lib/approval-inbox.ts`). It is the network surface of the human-in-the-loop Approval Gate, which `ApprovalGate` uses to require explicit human approval for high-risk operations. The HTTP server it exposes has no authentication of any kind and sets `Access-Control-Allow-Origin: *` on every route, including the state-changing `POST /approvals/:id/approve` and `/deny`. As a result, in versions 5.0.0 through 5.12.1, any party who can send an HTTP request to the inbox port — a co-located process, a container/SSRF on the same host, a remote client when the operator binds a non-loopback address, or any website the operator visits in a browser (via the wildcard CORS) — can enumerate pending approvals and approve them, defeating the entire human-in-the-loop control and causing the gated high-risk action (e.g. a shell command the agent was holding for review) to execute without consent. This issue is fixed in v5.12.2. `ApprovalInbox` now accepts a `secret` option. When set, the mutating endpoints `POST /:id/approve` and `POST /:id/deny` require an `Authorization: Bearer ` header, validated in constant time with `crypto.timingSafeEqual`. `startServer()` already binds to `127.0.0.1` by default; operators exposing the inbox on a network must set a secret.
osv:CVE-2026-58482
Complete source record
{
"id": "1d48df222885b7b6d7b183f71904efbdc66dcc3ea397cd0096ab1ad68fefc9be",
"url": "https://osv.dev/vulnerability/CVE-2026-58482",
"method": "structured",
"payload": {
"kind": "advisory",
"title": "Network-AI: ApprovalInbox HTTP server has no authentication — anyone can approve pending agent actions",
"summary": "Network-AI, a TypeScript/Node.js multi-agent orchestrator, has a shipped, exported, documented feature called `ApprovalInbox` (`lib/approval-inbox.ts`). It is the network surface of the human-in-the-loop Approval Gate, which `ApprovalGate` uses to require explicit human approval for high-risk operations. The HTTP server it exposes has no authentication of any kind and sets `Access-Control-Allow-Origin: *` on every route, including the state-changing `POST /approvals/:id/approve` and `/deny`. As a result, in versions 5.0.0 through 5.12.1, any party who can send an HTTP request to the inbox port — a co-located process, a container/SSRF on the same host, a remote client when the operator binds a non-loopback address, or any website the operator visits in a browser (via the wildcard CORS) — can enumerate pending approvals and approve them, defeating the entire human-in-the-loop control and causing the gated high-risk action (e.g. a shell command the agent was holding for review) to execute without consent. This issue is fixed in v5.12.2. `ApprovalInbox` now accepts a `secret` option. When set, the mutating endpoints `POST /:id/approve` and `POST /:id/deny` require an `Authorization: Bearer ` header, validated in constant time with `crypto.timingSafeEqual`. `startServer()` already binds to `127.0.0.1` by default; operators exposing the inbox on a network must set a secret.",
"evidence": "osv:CVE-2026-58482",
"products": [],
"severity": "medium",
"structured": {
"id": "CVE-2026-58482",
"aliases": [
"CVE-2026-58482",
"GHSA-mxjx-28vx-xjjj"
],
"metrics": [
{
"type": "CVSS_V3",
"score": "CVSS:3.1/AV:N/AC:H/PR:N/UI:R/S:U/C:L/I:H/A:N"
}
],
"affected": [
{
"ranges": [
{
"repo": "https://github.com/jovancoding/network-ai",
"type": "GIT",
"events": [
{
"introduced": "c7891a948d643c22fb40ae36ac05174fd1bf0e5b"
},
{
"fixed": "a59c13a1f0ce0e8a0779a90343eef92fac5ab4c3"
}
],
"database_specific": {
"extracted_events": [
{
"range": [
{
"introduced": "5.0.0"
},
{
"fixed": "5.12.2"
}
],
"source": "AFFECTED_FIELD",
"strategy": "StringRangeExpression"
},
{
"range": [
{
"fixed": "a59c13a1f0ce0e8a0779a90343eef92fac5ab4c3"
}
],
"source": "REFERENCES_COMMIT"
},
{
"range": [
{
"fixed": "a59c13a1f0ce0e8a0779a90343eef92fac5ab4c3"
}
],
"source": "REFERENCES_TAG",
"original_tag": "v5.12.2"
},
{
"cpe": "cpe:2.3:a:network-ai:network-ai:*:*:*:*:*:*:*:*",
"range": [
{
"introduced": "5.0.0"
},
{
"fixed": "5.12.2"
}
],
"source": "CPE_RANGE"
}
]
}
}
],
"versions": [
"v5.12.1",
"v5.12.0",
"v5.11.0",
"v5.10.2",
"v5.10.1",
"v5.10.0",
"v5.9.1",
"v5.9.0",
"v5.8.9",
"v5.8.8",
"v5.8.7",
"v5.8.6",
"v5.8.5",
"v5.8.4",
"v5.8.3",
"v5.8.2",
"v5.8.1",
"v5.8.0",
"v5.7.2",
"v5.7.1",
"v5.7.0",
"v5.6.1",
"v5.6.0",
"v5.5.9",
"v5.5.8",
"v5.5.7",
"v5.5.6",
"v5.5.5",
"v5.5.4",
"v5.5.3",
"v5.5.2",
"v5.5.1",
"v5.5.0",
"v5.4.5",
"v5.4.4",
"v5.4.3",
"v5.4.2",
"v5.4.1",
"v5.4.0",
"v5.3.2",
"v5.3.1",
"v5.3.0",
"v5.2.2",
"v5.2.1",
"v5.2.0",
"v5.1.4",
"v5.1.3",
"v5.1.2",
"v5.1.1",
"v5.1.0",
"v5.0.0"
],
"database_specific": {
"source": "https://storage.googleapis.com/cve-osv-conversion/osv-output/CVE-2026-58482.json"
}
}
],
"modified": "2026-10-28T02:30:52.974912212Z",
"provider": "osv",
"references": [
"https://github.com/Jovancoding/Network-AI/releases/tag/v5.12.2",
"https://github.com/CVEProject/cvelistV5/tree/main/cves/2026/58xxx/CVE-2026-58482.json",
"https://github.com/Jovancoding/Network-AI/security/advisories/GHSA-mxjx-28vx-xjjj",
"https://nvd.nist.gov/vuln/detail/CVE-2026-58482",
"https://github.com/Jovancoding/Network-AI/commit/a59c13a1f0ce0e8a0779a90343eef92fac5ab4c3"
],
"provider_metadata": {
"cwe_ids": [
"CWE-352",
"CWE-862"
],
"cna_assigner": "GitHub_M",
"osv_generated_from": "https://github.com/CVEProject/cvelistV5/tree/main/cves/2026/58xxx/CVE-2026-58482.json"
}
},
"related_cves": [
{
"value": "CVE-2026-58482",
"evidence": "/id"
}
],
"security_relevant": true,
"severity_evidence": "/severity/0/score"
},
"source_id": "core-osv",
"document_id": 521879,
"published_at": "2026-07-20T16:38:55.576+00:00"
}