Skip to main content

COMPEL Glossary / supply-chain-attack-agentic

Supply-chain attack (agentic)

Compromise of a tool vendor, model provider, or upstream data source that affects the agent — such as a poisoned Model Context Protocol server, a compromised library dependency, or a tampered fine-tuning dataset.

What this means in practice

Extends OWASP LLM03 to agentic systems; defenses include SBOM, signed dependencies, and runtime tool-call anomaly detection.

Synonyms

agent supply-chain compromise , upstream-compromise attack

See also

  • Tool registry — An authoritative inventory of tools — with schemas, permissions, owners, deprecation state, and audit log — that an agent may call.
  • Model Context Protocol (MCP) — An open protocol published by Anthropic in November 2024 for interoperability between models and tools — standardising how servers expose tools, resources, and prompts to AI clients.
  • Indirect prompt injection — Prompt injection delivered through content the model retrieves or ingests — emails, documents, web pages, or tool outputs — rather than through a direct user message.