Quickstart

Hierophant shows what a public GitHub repository would make your AI coding agent do, before you trust it. There is nothing to install and no account to create.

Check a repository

  1. Paste a URL like https://github.com/owner/repo into the scan form.
  2. Read the verdict and the Open Impact table for the agents you use.
  3. Open each finding. It shows the file, the line, why it matters, and how to fix it.
  4. Follow the steps under “Open this repo with less risk” before you click Trust.

Read the verdict

StatusMeaning
DangerousAt least one finding has critical or high severity.
Needs reviewAt least one finding has medium or low severity.
No issues foundThe rules in this ruleset version matched nothing. This is not a guarantee, and it does not prove a repository is harmless.

Read Open Impact

ImpactMeaning
Runs on openOfficial documentation says this runs automatically.
Runs after trust promptThis runs after you trust the folder or approve a prompt.
Steers the agentThis runs no command, but it changes the instructions your agent follows.
Behavior unknownThere is a finding, but official documentation does not explain the behavior.
Not detectedNo finding affects this agent in this ruleset version.

Read package checks

When a repo would download and run a package, the report shows what public registries and OSV say about it. A package listed as malicious makes the report Dangerous. A missing package, a known vulnerability, or a package created in the last 30 days makes it Needs review. No known issues does not mean a package is harmless.

Share a report

Each report has its own link. Anyone with the link can open it, and search engines are told not to index it. Code excerpts appear only in the browser tab that ran the scan, so they are not part of a shared link.

Add a badge

A report page includes a Markdown badge for your README. It shows the real status, the ruleset version, and the commit that was checked, and it links to the report. A badge describes that one commit, so scan again after you change agent files.

Local tools

A command-line scanner, a GitHub Action for CI, a read-only MCP server, and a command that locks your agent settings have been prepared. They use the same rules, but they have not been published yet, so they cannot be installed.