Skip to main content
Security

VigilSutra AI

Find leaked secrets and risky config before you push.

Offline, read-only auditor for source, CI, containers, Kubernetes, Terraform, and AI-agent setup. The CLI does not call the network.

  • Offline
  • Read-only
  • 42 rules
GitHub version
0.4.0
Licence
Apache-2.0
Language
TypeScript
Weekly downloads
1,542
GitHub stars
2

Live npm weekly downloads and the latest npm version, plus GitHub stars and the package.json version, fetched 2026-11-07 20:26 UTC. Cached 10 minutes. Weekly downloads cover 2026-10-31 to 2026-11-06, not a lifetime total. GitHub reported 3 open issues.

Use it when

  • 1

    You want a repo, CI, Docker, Kubernetes, or Terraform check that stays on your machine.

  • 2

    You need a CI gate such as --fail-on high.

  • 3

    You want reports with token-shaped text redacted.

What it does, and what it does not

Read this before the guide. A limit here is a limit of the tool.

Does

  • 42 rules, each with a bad case and a clean case, as stated on the project page.
  • Reads git history (SHA-1 or SHA-256) until a time limit, without running git.
  • Redacts token-shaped text in text, JSON, Markdown, SARIF, and HTML reports.
  • Leaves an untracked .env unread when a git index exists.
  • Writes vigilsutra-report/report.html and report.md.

Does not

  • Advisory matches use a snapshot from 2026-11-07, not a live feed.
  • Name tracking stops at object fields, dynamic imports, and function returns.
  • No mobile, runtime, or compliance verdict.
  • A clean report is not proof the project is secure.

Claims above are from the project pages and are attributed as stated there. They are not independently verified here.

How a run moves

From the VigilSutra project page, version 0.4.0. The boxes are the run. The table is how to read the last box.

  1. 1. Input

    Your repository

    Source, CI, Docker, Kubernetes, Terraform, and AI-agent setup. Read-only. An untracked .env stays unread when a git index exists.

  2. 2. Check

    42 rules, offline

    The CLI does not call the network. Git history is read as SHA-1 or SHA-256 until a time limit, without running git.

  3. 3. Report

    Redacted files

    vigilsutra-report/report.html and report.md, plus JSON and SARIF. Token-shaped text is redacted in each copy.

  4. 4. Decision

    You still judge it

    --fail-on high can stop a pipeline. A clean report is not proof the project is secure.

How to read the result

Result It means It does not mean
A match A rule found a pattern in the files you gave it. Advisory matches use a snapshot dated 2026-11-07, not a live feed.
Clean report No rule fired on this scan. Not a mobile, runtime, or compliance verdict.
--fail-on high The process can stop when a high finding exists. It does not fix the finding. fix . only shows edits until you pass --apply.

Wording is from the project page or README and is attributed as stated there. It is not independently verified here.

Commands and notes

  • vigilsutra audit . --fail-on high CI gate.
  • vigilsutra audit . --since <rev> limit the scan to files that differ.
  • vigilsutra sbom . CycloneDX file.
  • vigilsutra fix . shows edits. --apply writes them.
  • vigilsutra init --hook scan the git index before commit.
  • vigilsutra adapters install cursor also claude, codex, gemini, copilot. The project page says the file copy is tested and the IDE is not.

Guide

How to use the result. GitHub is the source repository. npm is the install.

How to read a result

From the VigilSutra project page, version 0.4.0. The panels above are the rule list. This is how to use the report.

  • Run npx [email protected] audit . The report is written to vigilsutra-report/report.html and report.md.
  • Use --fail-on high when a pipeline should stop. Use --since <rev> when only the files that differ should be scanned.
  • Git history is read as SHA-1 or SHA-256 until a time limit, without running git. An untracked .env stays unread when a git index exists.
  • Token-shaped text is redacted in the text, JSON, Markdown, SARIF, and HTML copies.
  • Advisory matches use a snapshot dated 2026-11-07, not a live feed. Name tracking stops at object fields, dynamic imports, and function returns.
  • A clean report is not a mobile, runtime, or compliance verdict, and it is not proof the project is secure.

Questions

What does VigilSutra AI do?

It audits a repository offline for secrets, CI, Docker, Kubernetes, Terraform, code, and AI-agent risks, then writes redacted HTML, Markdown, JSON, and SARIF. The project page says the CLI never calls the network and needs no API key.

Does a clean VigilSutra report mean the project is secure?

No. The project page says a clean report is not proof the project is secure, and it does not give a mobile, runtime, or compliance verdict.

WhatsApp