# Sckit Supply Chain Worm Hits MemTensor npm & PyPI Packages (@memtensor/memos-cloud-openclaw-plugin, MemoryOS)

> Malicious releases of the npm package @memtensor/memos-cloud-openclaw-plugin (0.1.21, 0.1.23, 0.1.25) and the PyPI package MemoryOS 2.0.34 bundle and launch a platform-specific Go native binary called 'sckit'. Static analysis shows credential-harvesting patterns and worm-like publishing/CI logic; exfiltration, credential reuse and reachability of the propagation code were not confirmed.

- **Published:** 2026-11-14T17:22:06Z
- **Last reviewed:** 2026-11-14T17:22:06Z
- **Canonical:** https://intel.threadlinqs.com/threat/TL-2026-6100
- **Card image:** https://intel.threadlinqs.com/og/threat/TL-2026-6100.png
- **ID:** TL-2026-6100
- **Severity:** HIGH
- **Category:** SUPPLY_CHAIN
- **Status:** MONITORING
- **Detections:** 9 (1 shown free below) · **IOCs:** 28 (3 shown free below)

## Description

On 2026-09-23 (UTC) a series of npm versions of the OpenClaw AI-agent memory plugin @memtensor/memos-cloud-openclaw-plugin were published that differ from the clean 0.1.20 baseline (2026-08-03). Versions 0.1.21, 0.1.23 and 0.1.25 add lib/sckit.js and six platform-specific executables under .sckit/ (unpacked size grew from about 272 KB to about 43.6 MB); interleaved versions 0.1.22 and 0.1.24 restored clean contents in the examined artifacts. A GitHub issue (#173) reports that the malicious npm artifacts have no matching repository commits or tags, which points to a compromised publishing credential rather than a repository compromise. A PyPI release, MemoryOS 2.0.34 (import name memos), carries a similar payload.

The malicious npm versions have no install/preinstall/postinstall hook. Instead lib/sckit.js is called from the plugin's own logic at gateway startup and on every memory recall event. It starts the native binary as a detached child with ignored standard streams, an empty error handler, no deduplication or rate limiting, the full host environment, and (npm path) the user's prompt text in the SCKIT_EVENT_TEXT variable. A helper, collectStageZero(), maps NPM_TOKEN / NODE_AUTH_TOKEN into the child environment but is not called in the reproduced index.js. Version 0.1.25 adds lib/tls-trust.js and a bundled CA file (.sckit/ca-roots.pem) that sets SSL_CERT_FILE on Linux when no system CA bundle exists; it does not disable verification. In the PyPI package, memos/log.py triggers the payload via memos/_stage0.py after logging configuration succeeds, so a normal import of memos can reach it; the Python trigger passes no prompt text. memos/_pypi_bridge.sh only runs when invoked as /app/twine-upload.sh inside GitHub Actions with INPUT_PASSWORD set, then runs memos/_initial_ci_delivery.py, which verifies a signed index against an embedded Ed25519 key, downloads four SHA-256-checked assets, runs a pinned emitter and prints SCKIT_CI_RESULT_V2. Activation of this path and credential theft were not observed.

The decoded embedded configuration uses schema sckit.runtime.v1 with campaign profiles named cloud-openclaw-semi-nuclear (npm) and memos-semi-nuclear (PyPI), an inventory root of $HOME, state directories $HOME/.openclaw/.cache/runtime and $HOME/.memos/.cache/runtime, and three skyleen.fr front hosts per ecosystem using /config, /status and /batch paths. not_after values correspond to 2026-10-23 and are configured fields, not confirmed kill switches. Embedded regular expressions match AWS access key IDs, GitHub, GitLab, npm, PyPI and Hugging Face tokens, HashiCorp Vault, Slack, Stripe live and SendGrid keys, JWTs, and names associated with passwords, private keys, API keys, session cookies and connection strings. The Linux amd64 binary contains the module path supplychain.local/campaign/cmd/implant and function names such as recursivePublish, prepareRemoteNode, prepareRemotePython and prepareRemoteWorkflow, plus a GitHub Actions template running './%s/linux-amd64/sckit stage0 --config64 %q' on push. These strings suggest intended credential collection and spread across repositories and ecosystems but do not prove any workflow was installed or package published.

All findings are from static analysis; the binaries were not executed in the cited reports. No CVE, CVSS score or actor attribution is given in the sources. On 2026-10-30 the maintainers closed the GitHub issue, stating that 0.1.21, 0.1.23 and 0.1.25 were no longer in the registry and that latest was 0.2.0; token rotation and trusted-publishing status were not confirmed. Socket.dev lists the same three npm versions as containing malware. Severity HIGH is analyst judgment.

## MITRE ATT&CK

- T1583.001 Domains
- T1195.002 Compromise Software Supply Chain
- T1059 Command and Scripting Interpreter
- T1480 Execution Guardrails
- T1552.001 Credentials In Files
- T1528 Steal Application Access Token
- T1005 Data from Local System
- T1119 Automated Collection
- T1071.001 Web Protocols

## Free detection (Splunk SPL, Microsoft KQL, Sigma)

**Sckit Implant Execution via Poisoned npm/PyPI Package (MemTensor OpenClaw / MemoryOS) — Exploitation** (`TL-2026-6100-SPL-001`)

Detects the Go native 'sckit' implant being launched with the 'stage0 --config64' command line used by the launcher and the embedded GitHub Actions template, from a .sckit directory, or as a child of node/python (lib/sckit.js and memos/_stage0.py spawn it detached). Associated with malicious @memtensor/memos-cloud-openclaw-plugin 0.1.21/0.1.23/0.1.25 and MemoryOS 2.0.34 (TL-2026-6100). Behaviour-based and independent of hashes.

- **Languages:** Splunk SPL, Microsoft KQL, Sigma
- **ATT&CK:** T1195.002, T1059
- **Data sources:** sysmon, XmlWinEventLog:Microsoft-Windows-Sysmon/Operational, DeviceProcessEvents
- **Severity:** high
- **Confidence:** high

### Splunk SPL

```spl
index=sysmon sourcetype="XmlWinEventLog:Microsoft-Windows-Sysmon/Operational" EventCode=1 earliest=-24h
| eval img=lower(Image), cmd=lower(CommandLine), parent=lower(ParentImage)
| where match(img, "[^a-z0-9_-]sckit([.]exe)?$") AND (match(cmd, "stage0|--config64") OR match(img, "[.]sckit[^a-z0-9_-]") OR match(parent, "[^a-z0-9_-](node|python[0-9.]*)([.]exe)?$"))
| stats count min(_time) as first_seen max(_time) as last_seen values(CommandLine) as command_lines values(ParentImage) as parents by host, User, Image
| convert ctime(first_seen) ctime(last_seen)
| table first_seen last_seen host User Image parents command_lines count
```

### Microsoft KQL

```kql
DeviceProcessEvents
| where Timestamp > ago(24h)
| where FileName in~ ("sckit", "sckit.exe")
| where ProcessCommandLine has_any ("stage0", "--config64")
    or FolderPath contains ".sckit"
    or InitiatingProcessFileName in~ ("node", "node.exe", "python", "python3", "python.exe", "pythonw.exe")
| project Timestamp, DeviceName, AccountName, FileName, FolderPath, ProcessCommandLine, InitiatingProcessFileName, InitiatingProcessCommandLine, SHA256
| order by Timestamp desc
```

### Sigma

```yaml
title: Sckit Implant Execution via Poisoned npm/PyPI Package
id: 3f6c1e52-8a41-4d7b-9c20-6e1d5b7a0f11
status: experimental
description: Detects execution of the Go implant sckit (stage0 --config64), from a .sckit folder or launched by node or python, as dropped by malicious MemTensor npm and PyPI packages (TL-2026-6100).
references:
    - https://www.stepsecurity.io/blog/sckit-supply-chain-worm-hits-memtensor-npm-pypi-scopes
    - https://github.com/MemTensor/MemOS-Cloud-OpenClaw-Plugin/issues/173
author: Threadlinqs Team
date: 2026-11-14
tags:
    - attack.initial_access
    - attack.t1195.002
    - attack.execution
    - attack.t1059
logsource:
    category: process_creation
detection:
    selection_img:
        Image|endswith:
            - '/sckit'
            - '\sckit.exe'
    selection_cli:
        CommandLine|contains:
            - 'stage0'
            - '--config64'
    selection_dir:
        Image|contains:
            - '/.sckit/'
            - '\.sckit\'
    selection_parent:
        ParentImage|endswith:
            - '/node'
            - '\node.exe'
            - '/python'
            - '/python3'
            - '\python.exe'
    condition: selection_img and (selection_cli or selection_dir or selection_parent)
falsepositives:
    - Unrelated internal tool also named sckit, verify path and signer
level: high

```

**Known false positives:**

- Unrelated internal tool or binary legitimately named sckit launched by node/python
- Security researchers detonating the sckit samples in a lab

**Tuning notes:** Sysmon for Linux and Windows both emit EventCode=1; change index/sourcetype if your Sysmon data lands elsewhere. Drop the parent-process clause if node/python launching a binary named sckit is common in your estate; allowlist signed internal tools by SHA256.

### 8 more detection rules (Splunk SPL, Microsoft KQL, Sigma), unlock with TL Creds

- Malicious MemTensor npm/PyPI Package Artifacts Written to Disk ([indicator], .sckit, [indicator]) — Exploitation (locked: unlock with TL Creds)
- Sckit PyPI CI Delivery Helper in GitHub Actions (twine-upload.sh / _pypi_bridge.sh) — Post-Exploitation (locked: unlock with TL Creds)
- Sckit Endpoint Connections to skyleen.fr Front Hosts and [indicator] — C2 (locked: unlock with TL Creds)
- Sckit Implant Binary Hash Match (npm and PyPI Payloads, All Platforms) — Exploitation (locked: unlock with TL Creds)
- Sckit Implant Spawning Publishing and Source-Control Tooling (Worm Propagation) — Post-Exploitation (locked: unlock with TL Creds)
- Sckit C2 HTTP Requests to skyleen.fr (/config, /status, /batch, initial-ci-v2) — C2 (locked: unlock with TL Creds)
- Sckit Runtime State Directory and CA Bundle Creation — Post-Exploitation (locked: unlock with TL Creds)
- Sckit C2 Front Host skyleen.fr and [indicator] Network Activity — C2 (locked: unlock with TL Creds)

Sign up free to unlock the detection rules: https://threadlinqs.com/?modal=signup

## Sample indicators (3 of 28)

Defanged, TLP:CLEAR.

- IP address: `139.84.223[.]178`
- Domain: `8a8acaf167b3[.]skyleen[.]fr`
- File hash: `381ac6dc1715d9298fe81b2a53a11f7b7d78e361ee3a6619ad54f8c4b062cc18`

Full set of 28 indicators unlocks with TL Creds: https://threadlinqs.com/tl-creds

## Sources

- [Sckit Supply Chain Worm Hits MemTensor npm & PyPI Scopes (StepSecurity)](https://www.stepsecurity.io/blog/sckit-supply-chain-worm-hits-memtensor-npm-pypi-scopes)
- [MemTensor incident report - GitHub issue #173](https://github.com/MemTensor/MemOS-Cloud-OpenClaw-Plugin/issues/173)
- [Socket.dev - @memtensor/memos-cloud-openclaw-plugin package page](https://socket.dev/npm/package/@memtensor/memos-cloud-openclaw-plugin)
- [Socket.dev - memoryos PyPI package page](https://socket.dev/pypi/package/memoryos)
- [StepSecurity Threat Center documentation](https://docs.stepsecurity.io/artifact-security/threat-center)
- [StepSecurity Harden-Runner](https://github.com/step-security/harden-runner)

## Full data

The other 8 detection rules (Splunk SPL / Microsoft KQL / Sigma) unlock with TL Creds; a free account includes daily creds: https://threadlinqs.com/?modal=signup
Indicator values unlock with TL Creds, at a lower price on Red and Purple; AI agents can query rules and IOCs through the Threadlinqs MCP server (Purple): https://intel.threadlinqs.com/mcp

Canonical: https://intel.threadlinqs.com/threat/TL-2026-6100
