brew-openclaw-cli-cve-2026-27646
Vulnerability from osv_homebrew
Summary
Sandboxed requester sessions could reach host-side ACP session initialization through /acp spawn.
OpenClaw already blocked sessions_spawn({ runtime: "acp" }) from sandboxed sessions, but the slash-command path initialized ACP directly without applying the same host-runtime guard first.
Affected Packages / Versions
- npm package:
openclaw - Affected versions:
<= 2026.3.2 - Patched version:
>= 2026.3.7
Details
ACP sessions run on the host, not inside the OpenClaw sandbox. The direct ACP spawn path in src/agents/acp-spawn.ts already denied sandboxed requesters, but /acp spawn in src/auto-reply/reply/commands-acp/lifecycle.ts called initializeSession(...) without first applying the same restriction.
In affected versions, an already authorized sender in a sandboxed session could use /acp spawn to cross from sandboxed chat context into host-side ACP runtime initialization when ACP was enabled and a backend was available.
Fix Commit(s)
61000b8e4ded919ca1a825d4700db4cb3fdc56e3
Fix Details
The fix introduced a shared ACP runtime-policy guard in src/agents/acp-spawn.ts and reused it from the /acp spawn handler in src/auto-reply/reply/commands-acp/lifecycle.ts before any ACP backend initialization. Regression coverage was added in src/auto-reply/reply/commands-acp.test.ts to prove sandboxed /acp spawn requests are rejected early, while existing ACP spawn behavior for non-sandboxed sessions remains unchanged.
Release Process Note
Patched version is pre-set to 2026.3.7 so the advisory can be published once that npm release is available.
Thanks @tdjackey for reporting.
{
"affected": [
{
"ecosystem_specific": {
"fix": "bump",
"range_state": "fixed",
"upstream_fixed_in": "2026.3.7"
},
"package": {
"ecosystem": "Homebrew",
"name": "openclaw-cli",
"purl": "pkg:brew/openclaw-cli"
},
"ranges": [
{
"events": [
{
"introduced": "0"
},
{
"fixed": "2026.7.1"
}
],
"type": "ECOSYSTEM"
}
]
}
],
"database_specific": {
"confidence": "high",
"source": "matched",
"strategy": "registry",
"upstream_evidence": [
{
"ecosystem": "npm",
"key": "pkg:npm/openclaw@2026.9.4",
"name": "openclaw",
"strategy": "registry",
"subject_version": "2026.9.4"
}
]
},
"details": "### Summary\nSandboxed requester sessions could reach host-side ACP session initialization through `/acp spawn`.\n\nOpenClaw already blocked `sessions_spawn({ runtime: \"acp\" })` from sandboxed sessions, but the slash-command path initialized ACP directly without applying the same host-runtime guard first.\n\n### Affected Packages / Versions\n- npm package: `openclaw`\n- Affected versions: `\u003c= 2026.3.2`\n- Patched version: `\u003e= 2026.3.7`\n\n### Details\nACP sessions run on the host, not inside the OpenClaw sandbox. The direct ACP spawn path in `src/agents/acp-spawn.ts` already denied sandboxed requesters, but `/acp spawn` in `src/auto-reply/reply/commands-acp/lifecycle.ts` called `initializeSession(...)` without first applying the same restriction.\n\nIn affected versions, an already authorized sender in a sandboxed session could use `/acp spawn` to cross from sandboxed chat context into host-side ACP runtime initialization when ACP was enabled and a backend was available.\n\n### Fix Commit(s)\n- `61000b8e4ded919ca1a825d4700db4cb3fdc56e3`\n\n### Fix Details\nThe fix introduced a shared ACP runtime-policy guard in `src/agents/acp-spawn.ts` and reused it from the `/acp spawn` handler in `src/auto-reply/reply/commands-acp/lifecycle.ts` before any ACP backend initialization. Regression coverage was added in `src/auto-reply/reply/commands-acp.test.ts` to prove sandboxed `/acp spawn` requests are rejected early, while existing ACP spawn behavior for non-sandboxed sessions remains unchanged.\n\n### Release Process Note\nPatched version is pre-set to `2026.3.7` so the advisory can be published once that npm release is available.\n\nThanks @tdjackey for reporting.",
"id": "BREW-openclaw-cli-CVE-2026-27646",
"modified": "2026-09-12T09:47:35Z",
"published": "2026-08-13T17:21:32Z",
"references": [
{
"type": "WEB",
"url": "https://github.com/openclaw/openclaw/security/advisories/GHSA-9q36-67vc-rrwg"
},
{
"type": "ADVISORY",
"url": "https://nvd.nist.gov/vuln/detail/CVE-2026-27646"
},
{
"type": "WEB",
"url": "https://github.com/openclaw/openclaw/commit/61000b8e4ded919ca1a825d4700db4cb3fdc56e3"
},
{
"type": "PACKAGE",
"url": "https://github.com/openclaw/openclaw"
},
{
"type": "WEB",
"url": "https://github.com/openclaw/openclaw/releases/tag/v2026.3.7"
},
{
"type": "WEB",
"url": "https://vulncheck.com/advisories/openclaw-mar-sandbox-escape-via-acp-spawn-command"
}
],
"schema_version": "1.7.3",
"severity": [
{
"score": "CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:L/I:H/A:N",
"type": "CVSS_V3"
},
{
"score": "CVSS:4.0/AV:L/AC:L/AT:P/PR:L/UI:N/VC:L/VI:H/VA:N/SC:N/SI:N/SA:N",
"type": "CVSS_V4"
}
],
"summary": "OpenClaw: Sandboxed /acp spawn requests could initialize host ACP sessions",
"upstream": [
"GHSA-9q36-67vc-rrwg",
"CVE-2026-27646"
]
}
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.