30 ClawHub skills secretly turn AI agents into crypto swarm

The Register

Original Source

The Register

// Analysis

Thirty ClawHub skills, published by a single author, are currently being used to covertly co-opt AI agents into a cryptocurrency mining swarm. This operation, identified by Ax Sharma, research lead at agentic AI security outfit Manifold, and dubbed "ClawSwarm," achieves this without employing traditional malware or obtaining explicit user consent. The skills were found on ClawHub, a public registry and marketplace for OpenClaw agent capabilities.

// Why It Matters

Signal Intelligence Assessment

The discovery of "ClawSwarm" reveals a novel supply-chain attack vector within agentic AI ecosystems, where seemingly benign skills can be co-opted for undisclosed, malicious purposes without traditional malware.
This incident signals a critical shift in agentic AI security, requiring OpenClaw developers and platforms to move beyond traditional vulnerability patching towards comprehensive governance and transparency for third-party skill marketplaces.
Builders deploying OpenClaw agents must now factor in the silent resource drain and potential for reputational damage from installed skills, necessitating granular permissioning and audit trails for agent actions.

Analyzed 17:00 PT · Trace No. 003 · April 2026

Back to Trace No. 003 · April 2026