cyber.hendryadrian.com
Security News · stepsecurity.io

Step Security Adds AI Agent Skill Inventory to Dev Machine Guard

· Original article ↗

Summary

Step Security says Dev Machine Guard now inventories AI agent skills across developer machines, flagging executable content and showing provenance, installation details, and version differences to help security teams review and respond to risky skills.

Key points

  • The new Agent Skills dashboard page inventories skills across an organization's developer devices and identifies which agents can load them, their installation scope, source, and device count.
  • The scanner checks shared, agent-specific, and project-level skill directories and resolves common symlinked installations.
  • Skills are flagged for bundled scripts, agent hooks, or shell commands in SKILL.md; Step Security says the flags identify content worth review, not necessarily malicious skills.
  • The dashboard shows skill provenance and compares SKILL.md content hashes across devices to highlight differing or potentially outdated copies.
  • Per-device details include the device, user, agent, installation scope, and last-seen time; the inventory can be filtered and exported.
  • The article cites the ClawHavoc campaign and Snyk's ToxicSkills research as examples of risks from malicious or flawed AI agent skills.

Article Details

Event Type
Product feature announcement: Step Security added an Agent Skills inventory to Dev Machine Guard.
Impact
The feature gives security teams fleet-wide visibility into installed AI agent skills, their sources, the agents and devices that can load them, and whether they contain executable content. The article describes this as a way to identify affected machines and prioritize review during incidents. It cites prior ecosystem risks: the ClawHavoc campaign reportedly planted 1,184 malicious skills across 12 publisher accounts and delivered Atomic Stealer (AMOS); Snyk's audit of 3,984 public skills found that 36% contained security flaws, including active malicious payloads.

Malware

Atomic Stealer (AMOS)1,184 malicious skills across 12 publisher accounts on a public skill registry, delivering the Atomic Stealer (AMOS) malware that targets macOS crypto wallets, SSH keys, and browser credentials. At the peak of the

Products

Claude CodeIf your developers use Claude Code, Codex, GitHub Copilot, Cursor, OpenCode, Gemini CLI, Amp, Factory, or any other AI coding agent, there is a good chance skills are already installed on their machines. Until now, mostClineAI developer tool chain all year. The Miasma worm campaign explicitly targeted AI coding agents, and the Cline v2.3.0 compromise showed how an AI coding tool itself can become the delivery vehicle. Skills follow theCursorIf your developers use Claude Code, Codex, GitHub Copilot, Cursor, OpenCode, Gemini CLI, Amp, Factory, or any other AI coding agent, there is a good chance skills are already installed on their machines. Until now, mostDev Machine GuardDev Machine Guard now inventories AI agent skills installed on developer machines. A new Agent Skills page in the StepSecurity dashboard shows every skill detected across your fleet, which AI coding agents can load it,OpenClawand found that 36% contained security flaws, including active malicious payloads targeting Claude Code, OpenClaw, and Cursor users. The researchers demonstrated that three lines of markdown in a SKILL.md file were

Related Articles