Manifest Platform From Manifold Targets AI Agent Supply Chain Security Gaps

Why It Matters

By exposing hidden dependencies and malicious skill chains, Manifest helps enterprises mitigate AI‑driven attack surfaces that existing scanners overlook, strengthening overall AI operational security.

Key Takeaways

• Manifest maps AI agent dependencies and external system interactions
• Platform indexes over 100,000 AI assets for free access
• Report finds 238,000 unique AI skills, some malicious
• Detection tools vary widely; lack context leads to false positives
• Enterprise add‑on covers browser extensions and protocol servers

Pulse Analysis

The rapid expansion of AI agents—from chat assistants to autonomous workflow bots—has created a sprawling supply chain of skills, plugins, and model‑serving components. Each piece carries its own trust assumptions, and when combined, they can form unexpected attack vectors. Traditional security tools focus on static file analysis, which often fails to capture the relational risk that emerges when a benign skill calls a compromised API or when a malicious author injects code into a popular registry. Manifold’s Manifest tackles this gap by constructing execution graphs that trace what a skill does and environment graphs that map authorship, similarity, and cross‑registry relationships, delivering a holistic view of AI agent behavior.

Manifold’s research underscores the scale of the problem: more than 238,000 distinct AI skills exist across multiple marketplaces, and malicious examples have already been promoted and executed thousands of times, leading to identity data exfiltration. The study also reveals that existing detection solutions—static analyzers, LLM classifiers, and behavioral monitors—produce widely divergent results, reflecting a systemic lack of contextual awareness. By integrating ecosystem‑level data such as author activity, dependency chains, and infrastructure connections, Manifest improves signal quality, reduces false positives, and offers a unified risk baseline for security teams.

For enterprises, the availability of a free, open‑access platform with a searchable database of over 100,000 assets lowers the barrier to entry for AI supply‑chain risk management. The platform’s enterprise extensions, covering browser extensions and model‑context protocol servers, further extend protection to the broader AI‑driven stack. Backed by an $8 million Series A round, Manifold is positioned to become a critical player in securing the next generation of AI operations, where visibility into inter‑component relationships will be as essential as traditional perimeter defenses.