Skills Aggregation and Distribution via npm Package
This rule flags newly published npm packages whose description, keywords, or README mention two or more AI agent platforms (like Claude, Copilot, Gemini) together with language about aggregating, collecting, or reselling their skills or capabilities. That combination is the textual fingerprint of a package built to harvest proprietary agent content from multiple platforms and redistribute it without permission.
How the attack works
An attacker (or an automated scraper) collects skill definitions, plugins, or capability files from one or more AI agent ecosystems. They package this content into an npm module and write metadata that markets it as a directory, marketplace, or aggregator spanning multiple AI platforms. The package is published to the public npm registry, making the harvested content available for anyone to install and reuse. This exposes proprietary or platform-restricted skill content outside its intended ecosystem and can violate the terms of the source platforms.
Netzilo detection
Netzilo reports this behaviour when it is observed.
- Signature ID
- b2b3d47a-9f61-4b8b-8e0a-3d6f1c9a5e02
- Severity
- Medium
Why it matters
Organizations that built proprietary agent skills or plugins can have that work copied, repackaged, and distributed through an unrelated channel, undermining licensing controls and giving competitors or attackers easy access to capability logic they didn't develop.
What you can do
- →Audit npm packages your teams install for AI-agent tooling; check whether any repackage skill/capability files sourced from platforms you use.
- →Review your own published packages' descriptions and READMEs to ensure they don't unintentionally look like aggregation services if they're just multi-platform SDKs.
- →Restrict or license-gate distribution of internal skill or plugin definitions so they can't be scraped and republished.
- →Monitor for public packages referencing your platform's skill format alongside competitor platforms, which may indicate unauthorized aggregation.
Known benign look-alikes
- Legitimate SDKs or developer tools that advertise compatibility with multiple AI coding assistants (e.g. "works with Claude, Copilot, and Gemini") without redistributing any platform's proprietary skill or capability content.
- Documentation, comparison, or benchmarking packages whose README or keywords mention several AI platforms for informational purposes only, with no aggregation or marketplace language present.
- Internal or company packages whose description simply lists supported AI assistant integrations rather than describing a skill or capability aggregation/harvesting service.