23 ClawHub Plugins Published Under Official @openclaw/ and @clawhub/ Scopes by Unauthorized Accounts, Exposing Claude Code and Cursor to Supply Chain Risk

Manifold Security found 23 ClawHub plugins published under @openclaw/ and @clawhub/ scopes by unauthorized accounts, exposing Claude Code, Cursor, and Codex to silent supply chain compromise.

ClawHubTool execution / MCPRegistry scope squatting / supply chain impersonationClawHub plugin registry — primary source for Claude Code, Cursor, and Codex plugin installs (1,500+ plugins)

What happened

Third-party accounts published 23 code-executing plugins under @openclaw/ and @clawhub/ organizational scopes on ClawHub without authorization, making them appear to be official OpenClaw tools to any developer browsing or scripting plugin installs.

Why it matters

23 plugins with user-level authority inside agent environments appeared as trusted official tools on the registry used by Claude Code, Cursor, and Codex. Developers installing them had no visible signal the plugins were not official. Several performed high-privilege actions including autonomous payment processing, host-level git commands, agent configuration export, and external API connections.

Missing authorization check

Not applicable: no agent authorization boundary was crossed in this incident.

Would PP block it?

The compromise ran through package, credential, or vendor infrastructure rather than through an agent tool call, so there is no agent action for an authorization gate to hold.

Incident analysis

Timeline and technical read

Timeline

2026-06-17
Manifold Security reports 23 scope-squatting plugins to ClawHub via GitHub’s security advisory workflow.
2026-06-18
Manifold sends a courtesy email to ClawHub to confirm the disclosure was received.
2026-06-19
ClawHub unlists all 23 misleading plugins and adds a formal dispute process for reporting unauthorized namespace usage.
2026-06-22
Manifold Security publishes findings; CybersecurityNews and Help Net Security cover the disclosure.

Technical breakdown

ClawHub’s scoping system mirrors npm’s @owner/ convention, where the prefix signals the publishing organization. Unlike npm, ClawHub did not consistently enforce that only verified org members could publish under reserved scopes.
Of 1,508 plugins in the catalog, 557 carry an @owner/ prefix, but not all have verified ownership. The 23 rogue plugins belong to 15 distinct external accounts.
Plugin names like @openclaw/security-gate, @openclaw/fiat-wallet, and @clawhub/aisa-twitter-api impersonated platform-level tools. @openclaw/security-gate — a security-review plugin — even passed ClawHub’s own scanner despite not belonging to OpenClaw.
All 23 plugins execute code inside agent environments. Several perform high-privilege actions including autonomous payment processing, host-level git commands, agent config export, and external API connections.
Manifold found no confirmed malicious code in reviewed versions, but the supply chain risk is latent: a single future update to any of these plugins could introduce harmful behavior with no visible warning to developers who trust the @openclaw/ scope.

Authorization boundary

Where the authorization boundary should have been

This incident is categorized as Tool execution / MCP. The relevant Permission Protocol gate is Runtime Gate. The read is conditional: the block only applies where the real action boundary is routed through a gate.

If enforced at: Payment gate, data mutation gate, git execution gate, external API call gate
Still needs: Plugin installation path, scope validation at the registry level, and low-privilege actions below PP’s enforcement boundary
Receipt required for: Autonomous payment processing, git operations, agent configuration exports, and external API connections made by the plugin inside the agent

No agent took an action in this incident. Permission Protocol gates what an agent does, so it does not apply where the harm required no agent action.

Lessons for teams

Registry scope enforcement must happen at publication, not post-publication audit — a plugin already indexed under a trusted scope can be installed by thousands of developers before an audit catches it.
Official-looking scopes are a high-trust signal that attackers can exploit; developers should cross-reference the publishing account with the organization’s known contributors before installing any plugin.
A clean scanner verdict on a scope-squatted plugin provides false assurance: @openclaw/security-gate passed ClawHub’s own audit despite not belonging to OpenClaw.
AI agent plugin registries should enforce scope ownership at publish time, not rely on reporting workflows activated after discovery — the disclosure-to-unlist window is an exposure window.
No confirmed malicious code today does not mean the supply chain is safe: a single update to any of the 23 plugins could weaponize the existing installation base without any new trust signals.

Related incidents and controls

Critical2026-02-05

ClawHavoc: 1,184 Malicious Skills Uploaded to OpenClaw ClawHub Marketplace Delivering Atomic Stealer Targeting API Keys, SSH Creds, and Browser Passwords

High2026-06-23

Fake AI Agent Skill Bypassed All Security Scanners and Reportedly Reached 26,000 Agents via Staged External Payload

Critical2026-02-17

Clinejection: Cline CLI 2.3.0 npm Supply Chain Attack Silently Installed OpenClaw on ~4,000 Developer Machines via Prompt Injection and GitHub Actions Cache Poisoning

Critical2026-06-16

Sapphire Sleet (North Korean APT BlueNoroff) Hijacked Mastra npm Scope and Republished 142 Packages Injecting easy-day-js RAT Over 88 Minutes

High2026-05-19

TrapDoor Cross-Ecosystem Supply Chain: 34 Packages Planted Hidden AI Instructions in CLAUDE.md and .cursorrules to Poison Coding Assistants on Developer Machines

AI agent compliance controls Signed audit evidence

Start small

Put the relevant gate at this action boundary.

This incident maps to Runtime Gate. Start with the boundary that controls the actual action, then require a signed receipt before execution.

Replay this incident with a signer in the loop