Checks
Besides full repositories, Hierophant can check individual things you add to an agent. Open the check page and choose a type. Nothing is installed or executed, and each result can be shown as a badge.
Packages
For npm and PyPI packages, Hierophant reads the registry record and looks up the version in OSV. It flags packages listed as malicious, missing names, known vulnerabilities, and packages created in the last 30 days. It also notes install scripts and deprecation.
MCP servers
For servers in the official MCP Registry, Hierophant inspects the launch command, remote hosts, declared secrets, npm and PyPI packages, and the description. It does not run the server or inspect what its tools do at runtime.
Agent skills
For a public GitHub folder that contains SKILL.md, Hierophant reads up to 25 text files at one commit. It looks for hidden instructions, remote code, concealed payloads, and allowed-tools settings that pre-approve any command.
Token metadata
This is not a safety rating, audit, or investment advice. Hierophant reads a token’s on-chain name and symbol, plus the metadata URI on Solana, and looks for hidden characters or instructions aimed at AI agents.
It does not check the contract, liquidity, holders, ownership, price, or whether a token can be sold. A token with no injection found can still be a scam. Off-chain metadata is never fetched.
Supported chains: Ethereum, Base, Arbitrum, Optimism, Polygon, BNB Chain, and Solana.
Results and badges
Results are computed when requested and are not stored. Badges show the current status, can be cached for up to an hour, and are not a guarantee.