Curated advisory
MCP tool description poisoning
How misleading MCP descriptions can redirect agent behavior.
Start with a plain-language summary before the technical trace.
Use local Guard checks for tool descriptions, secret reads, and risky commands.
Connect Guard Cloud when the pattern should be remembered across a team.
Related explainers
Follow the concepts behind this advisory.
prompt-injection
Prompt injection
A prompt injection is a trick note for an AI tool.
Open guidemcp-tool-poisoning
MCP security
MCP lets agents use tools. Bad tools can lie or overreach.
Open guidesupply-chain
Skill and plugin safety
Agent skills are tiny helpers. Some helpers ask for too much.
Open guideprompt-injection
Community warnings
Learn from blocked actions without exposing private work.
Open guide