Skip to content
Back to search

vbsec

Security scanning skill for Claude Code, Codex and Antigravity. Finds the 21 most common vulnerabilities in AI-written code and shows how to fix each one.

Safety scan

A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.

Checked today

Check it deeper on your machine

It has changed since its scan on 11 Nov 2026.

As a Claude Code plugin, install it and then ask Claude Code:

/plugin marketplace add keithmackay/mackayi/plugin install scanskillsafety@mackayiDo a deep scan of https://github.com/tanviet12/vbsec with scanskillsafety

Or from a terminal, clone it and run it:

git clone https://github.com/keithmackay/scanskillsafety-skill ~/.claude/skills/scanskillsafetynode ~/.claude/skills/scanskillsafety/cli.cjs --deep 'https://github.com/tanviet12/vbsec'

It checks the exact commit: every file, its packages against OSV (api.osv.dev), how it compares with this page's scan (findsafeskills.com receives the repository's address), and, if you agree, Claude reading every file with your own Claude subscription. It never runs the code.

Badge for your README
findsafeskills safety rating

Markdown

[![findsafeskills safety rating](/api/listings/tanviet12--vbsec/badge.svg)](/listings/tanviet12--vbsec)

HTML

<a href="/listings/tanviet12--vbsec"><img src="/api/listings/tanviet12--vbsec/badge.svg" alt="findsafeskills safety rating" /></a>

Install

i

Install instructions from the README

As written by the repo's author. findsafeskills has not verified these steps.

You need one of: [Claude Code](https://docs.claude.com/claude-code), [OpenAI Codex CLI](https://developers.openai.com/codex), [Google Antigravity](https://antigravity.google).

git clone https://github.com/tanviet12/vbsec ~/vbsec
~/vbsec/scripts/install.sh

The script detects which of these tools you have and installs the skill for each. To update later: `cd ~/vbsec && git pull`.

Manual install, installing for one specific tool, troubleshooting: [docs/en/installation.md](docs/en/installation.md).

What was scanned

The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.

README.md · README · 261 lines

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

This file has changed since it was scanned, so this is not exactly the text that was scanned, and no line is marked. The findings are listed with the safety scan above.

  1. <h1 align="center">vbsec — Security scanning for AI-written code</h1>
  2. <p align="center">A security scanning skill for Claude Code, Codex and Antigravity: it reads your code, finds 30 vulnerability types, from classic injection bugs to the logic flaws of multi-tenant SaaS and AI agents, and shows you how to fix each one.<br>Free · Open source · Nothing extra to install</p>
  3. <p align="center">
  4. <a href="./LICENSE"><img src="https://img.shields.io/badge/license-MIT-blue" alt="MIT license"></a>
  5. <a href="https://github.com/tanviet12/vbsec/stargazers"><img src="https://img.shields.io/github/stars/tanviet12/vbsec?style=flat&color=2ea44f" alt="GitHub stars"></a>
  6. <img src="https://img.shields.io/badge/Claude%20Code-%E2%9C%93-d97757" alt="Claude Code">
  7. <img src="https://img.shields.io/badge/OpenAI%20Codex-%E2%9C%93-black" alt="OpenAI Codex">
  8. <img src="https://img.shields.io/badge/Google%20Antigravity-%E2%9C%93-4285f4" alt="Google Antigravity">
  9. </p>
  10. <p align="center">
  11. <b><a href="#install">Install</a></b> ·
  12. <b><a href="docs/examples/sample-report.md">Sample report</a></b> ·
  13. <b><a href="#what-vbsec-catches">30 vulnerability types</a></b> ·
  14. <b><a href="README.vi.md">🇻🇳 Tiếng Việt</a></b>
  15. </p>
  16. <p align="center">
  17. <a href="docs/examples/sample-report.md"><img src="docs/images/sample-report.png" alt="vbsec report: verdict FAIL, table of 6 critical issues" width="760"></a><br>
  18. <sub>A real report from scanning a sample Express + React app. Click the image to read the full report.</sub>
  19. </p>
  20. ## What vbsec does
  21. - **Finds vulnerabilities in AI-written code**: passwords in source, SQL injection, broken access control, JWTs accepted without signature checks, wide-open CORS and more: 21 classic types
  22. - **Catches the bugs of SaaS apps and AI agents**: one company reading another's data, a side path (restore, import, copy, agent tool) skipping the permission check the main screen has, closed periods edited, money counted twice, prompt injection through user-written text, agent tools that act without a real human confirmation: 9 more types
  23. - **Explains them for people who aren't security experts**: every issue says why it's dangerous, how an attacker would exploit it step by step, the current code, and the fixed code
  24. - **Understands code instead of matching strings**: it traces data from user input to the dangerous call, so it raises far fewer false alarms than pattern-based scanners
  25. - **Goes deep on 6 languages**: Go, PHP, TypeScript/JavaScript, Python, .NET/C#, Dart/Flutter, including popular frameworks such as Express, NestJS, Next.js, React, Django, FastAPI, Laravel, ASP.NET Core and Flutter mobile apps
  26. - **Scans exactly what you need**: the whole repo, uncommitted changes only, one commit, one pull request, the last N days of commits, or a commit range such as `v1.2.0..HEAD` before a release
  27. - **Works without git**: drop AI-generated code into a folder and scan it right away
  28. - **Stays fast on big repos**: splits the work across several agents running in parallel, then merges the results
  29. - **Reports in English or Vietnamese**, saved as a file in `vbsec-reports/` you can hand to whoever fixes it, with a JSON summary at the end for CI/CD
  30. ## Example: what vbsec catches
  31. The code below works. Clicking through the app by hand, you'd never notice a problem. AI assistants write code like this all the time.
  32. ```typescript
  33. // src/lib/auth.ts
  34. const payload = jwt.decode(token); // decodes only, does NOT verify the signature
  35. (req as any).user = payload;
  36. // src/routes/search.ts
  37. const rows = await sequelize.query(
  38. `SELECT id, name, price FROM products WHERE name LIKE '%${q}%'` // q comes from the URL
  39. );
  40. ```
  41. vbsec reports 2 **CRITICAL** issues:
  42. | Issue | Impact | Fix vbsec suggests |
  43. |---|---|---|
  44. | `JWT-NONE-ALGORITHM` | Anyone can mint their own `role: "admin"` token without the secret key | Use `jwt.verify(token, secret, { algorithms: ["HS256"] })` |
  45. | `SQL-INJECTION` | Typing `' UNION SELECT email, password_hash...` into the search box dumps the whole user table | Pass parameters with `` replacements: { q: `%${q}%` } `` instead of building the string |
  46. Read the [full sample report](docs/examples/sample-report.md) to see how vbsec explains each issue.
  47. ## Test results
  48. The repo ships sample apps with known, planted bugs. Each set also has traps that look like bugs but are safe, to measure false alarms.
  49. - **9 classic sets** (Go, PHP, TypeScript, Python, .NET, Flutter), including hard cases: data flowing through several files, sanitizers written wrong, race conditions when deducting a balance.
  50. - **4 SaaS sets, new in v0.8**: `saas-hard` (MCP server, over-privileged agent, tenant leaks, permission checks that differ between paths, over-exposed responses), `ledger-hard` (money and state rules, concurrency, logs, sessions), `misc-hard` (zip bombs, PII in logs, open redirect), `lessons-hard` (one case per real-world security lesson). Each comes with a sample project rules file in `tests/context/`.
  51. | | Latest run, 9 classic sets (v0.7) |
  52. |---|---|
  53. | Bugs caught | **54/55** |
  54. | False alarms on safe traps | **0** |
  55. v0.8 results, measured 2026-11-16. **Every column uses the same model, Claude Opus 5.5**; the only difference is whether vbsec is used. One run per configuration.
  56. **1. Real code** — a production multi-tenant SaaS codebase with an MCP server for an AI assistant, about 90 Go files. The answer key is 25 High / Medium issues found earlier by in-depth security reviews (several passes, with design docs); vbsec re-scanned the exact code that still had them.
  57. | | Opus, no skill | vbsec v0.7 | Opus + vbsec v0.8 (with `--context`) |
  58. |---|---|---|---|
  59. | High issues (5) | 0 | 0 | **5** |
  60. | Medium issues (20) | 1 | 0 | **16.5** |
  61. | **Total (25)** | **1 (4%)** | **0** \* | **21.5 (86%)** |
  62. \* v0.7: two release-gate scans over the same period caught none of these High / Medium issues.
  63. Without vbsec, the model read all the code and concluded "no serious issues"; it missed prompt injection through MCP, tools where the AI fills in its own confirmation, permissions leaking through a copy path, and writes into already-closed records. These are the classes vbsec v0.8 was built to catch.
  64. **2. Test sets in this repo** (4 new SaaS sets, 61 planted bugs, each within 1–3 files):
  65. | Set (planted bugs) | Opus, no skill | Opus + vbsec v0.7 | Opus + vbsec v0.8 |
  66. |---|---|---|---|
  67. | saas-hard (17) | 17 | 11 | 17 (16 with `--context`) |
  68. | ledger-hard (13) | 12 | 7 | 12 |
  69. | misc-hard (3) | 3 | 0 | 3 (with `--context`) |
  70. | lessons-hard (28) | 15–18 | 10 | 19 |
  71. | **Total (61)** | **47–50** | **28** | **51** |
  72. The sets are small and each bug sits close to its cause, so a strong model with no skill finds many of them too; v0.8 still scores highest and nearly doubles v0.7. Beyond the count, vbsec gives a saved, standard report (severity, fix, JSON for CI) instead of a chat answer. The real gap shows on larger multi-layer code, as in part 1.
  73. <!-- v0.8 bench -->
  74. AI output can vary between runs. Re-run it yourself with `./scripts/run-fixtures.sh` (see [`tests/README.md`](tests/README.md)). vbsec has also been tried on OWASP Juice Shop and caught the vulnerability classes matching its documented challenges.
  75. ## How vbsec works
  76. **It reads code like a reviewer, not like grep.** Many scanners flag every `query(` as "possible SQL injection", so they cry wolf constantly and people stop reading. vbsec reads the surrounding code, works out where the data comes from, and only then decides.
  77. Take the same line of code:
  78. ```typescript
  79. db.query(`SELECT * FROM products WHERE name = '${x}'`)
  80. ```
  81. - If `x` comes from a search box the user types into (`req.query.q`): **flagged**, because anyone can type malicious SQL there.
  82. - If `x` is a constant in the code or a value from a config file: **not flagged**, because outsiders can't change it.
  83. To tell these apart, vbsec sorts data into 4 trust levels, from "sent by a stranger" (forms, URLs, headers, uploads) to "produced by the system" (constants, environment variables). Only stranger-supplied data that reaches a dangerous call without being cleaned gets reported.
  84. **It knows each language and framework.** On top of the general rules, vbsec has dedicated rules for Go, PHP, TypeScript/JavaScript, Python, .NET and Dart/Flutter, so it knows each library's traps. For example: `Prisma.sql` is safe but `$queryRawUnsafe` is not, `bypassSecurityTrustHtml` in Angular switches off XSS protection, Gin's debug mode left on in production, and a Flutter app keeping its login token in `SharedPreferences` instead of the Keychain.
  85. **Small repos scan instantly, big ones get split up.** Under about 20 code files, vbsec scans directly in roughly 30–60 seconds. Larger repos are split into parts for up to 3 agents working in parallel, then the results are merged and de-duplicated.
  86. **One issue per line item.** A line of code with 2 problems (say, reading another user's data and a race condition) shows up as 2 separate issues. Counts stay honest, and each one can be ticked off as it's fixed.
  87. **Reports in English or Vietnamese.** Vietnamese is the default; add `lang=en` for English. The report ends with a fixed-format JSON block in English, so CI/CD can read it and block merges while critical issues remain.
  88. **The same rules on all three tools.** Claude Code, Codex and Antigravity share one rule set and produce the same findings. The only difference: Claude Code runs in parallel, so it's faster on large repos.
  89. ## Install
  90. You need one of: [Claude Code](https://docs.claude.com/claude-code), [OpenAI Codex CLI](https://developers.openai.com/codex), [Google Antigravity](https://antigravity.google).
  91. ```bash
  92. git clone https://github.com/tanviet12/vbsec ~/vbsec
  93. ~/vbsec/scripts/install.sh
  94. ```
  95. The script detects which of these tools you have and installs the skill for each. To update later: `cd ~/vbsec && git pull`.
  96. Manual install, installing for one specific tool, troubleshooting: [docs/en/installation.md](docs/en/installation.md).
  97. ## Usage
  98. | Tool | How to run |
  99. |---|---|
  100. | Claude Code | `/vbs-scan-security` |
  101. | OpenAI Codex CLI | `$vbs-scan-security` |
  102. | Google Antigravity | say "scan security for this repo" |
  103. ```bash
  104. /vbs-scan-security lang=en # scan the whole folder (default), English report
  105. /vbs-scan-security uncommitted lang=en # uncommitted changes only; worth running before every commit
  106. /vbs-scan-security pr id 42 lang=en # scan pull request #42
  107. /vbs-scan-security commit within 7days lang=en # commits from the last 7 days
  108. /vbs-scan-security range v1.2.0..HEAD lang=en # everything changed since the last release
  109. ```
  110. **Tell vbsec your business rules.** Logic bugs (a closed period still editable, a permission only the owner may grant) can't be seen from code alone. Write your rules in `.vbsec/context.md` (loaded automatically) or pass `--context <file>`. Template: [`project-context.md`](skills/vbs-scan-security/references/project-context.md).
  111. Every report prints the vbsec version it ran (`vbsec_version`) and a coverage table: files read in full, files only pattern-searched, rules not loaded. `--auto-fix` only applies mechanical fixes (parameterized SQL, output escaping, turning off debug mode, dependency bumps); permission, logic and AI fixes stay as suggested patches for a human to review.
  112. Reports are saved to `vbsec-reports/scan-<timestamp>.md` inside the scanned folder. Add `vbsec-reports/` to `.gitignore`; vbsec reminds you if it's missing.
  113. All options: [docs/en/usage.md](docs/en/usage.md).
  114. ## What vbsec catches
  115. | Category | Issues |
  116. |---|---|
  117. | Leaked secrets, config | `HARDCODED-SECRET` · `VERBOSE-ERROR-DEBUG-MODE` · `CORS-MISCONFIG` |
  118. | Injection | `SQL-INJECTION` · `COMMAND-INJECTION` · `XSS` · `INSECURE-DESERIALIZATION` · `SSRF` · `PATH-TRAVERSAL` |
  119. | Authentication, authorization | `BROKEN-ACCESS-CONTROL` · `IDOR` · `MASS-ASSIGNMENT` · `JWT-NONE-ALGORITHM` · `WEAK-PASSWORD-HASHING` · `CSRF` |
  120. | Abuse protection | `BRUTE-FORCE` · `MISSING-RATE-LIMIT` · `RACE-CONDITION` · `UNRESTRICTED-FILE-UPLOAD` |
  121. | Dependencies | `OUTDATED-DEPENDENCY` · `SLOPSQUATTING` (package names the AI made up, registered first by attackers) |
  122. | AI, MCP | `LLM-PROMPT-INJECTION` · `EXCESSIVE-AGENCY` |
  123. | SaaS permissions, data | `TENANT-ISOLATION` · `AUTHZ-PATH-PARITY` · `EXCESSIVE-DATA-EXPOSURE` · `BUSINESS-STATE-INTEGRITY` |
  124. | Logs, sessions, cost | `SENSITIVE-DATA-IN-LOGS` · `AUTH-SESSION-LIFECYCLE` · `RESOURCE-EXHAUSTION` |
  125. Plus `VULNERABLE-DEPENDENCY` (live CVE lookup, only with `--sca`). A real, exploitable bug outside these 30 types is still reported as `OTHER-CWE-<n>` (open redirect, XXE, SSTI...) and counts toward the verdict. The AI/MCP and tenant rules only load when the repo shows signs of them (an LLM/MCP SDK, a `tenant_id`-style column).
  126. Severity, vulnerable and safe code examples, and which languages have dedicated rules: [docs/en/rules.md](docs/en/rules.md).
  127. ## Limitations
  128. vbsec is a first line of defense, not proof that your system is secure.
  129. - It does not replace a security review by professionals
  130. - It does not guarantee catching 100% of vulnerabilities
  131. - It does not query CVE databases online unless you add `--sca`. Without it, check dependencies with `npm audit`, `pip-audit`, `govulncheck`, `composer audit`
  132. - Business-logic checks are only as good as the rules it knows: without a project rules file, it infers them from code and says so
  133. ## Add a badge to your repo
  134. Scanned with vbsec and fixed everything? Add this badge to your README:
  135. [![vbsec: security scanned · passed](https://img.shields.io/badge/vbsec-security%20scanned%20%C2%B7%20passed-2ea44f?logo=data%3Aimage%2Fsvg%2Bxml%3Bbase64%2CPHN2ZyB4bWxucz0iaHR0cDovL3d3dy53My5vcmcvMjAwMC9zdmciIHZpZXdCb3g9IjAgMCAyNCAyNCIgZmlsbD0ibm9uZSIgc3Ryb2tlPSIjZmZmIiBzdHJva2Utd2lkdGg9IjIuMiIgc3Ryb2tlLWxpbmVjYXA9InJvdW5kIiBzdHJva2UtbGluZWpvaW49InJvdW5kIj48cGF0aCBkPSJNMjAgMTNjMCA1LTMuNSA3LjUtNy42NiA4Ljk1YTEgMSAwIDAgMS0uNjctLjAxQzcuNSAyMC41IDQgMTggNCAxM1Y2YTEgMSAwIDAgMSAxLTFjMiAwIDQuNS0xLjIgNi4yNC0yLjcyYTEuMTcgMS4xNyAwIDAgMSAxLjUyIDBDMTQuNTEgMy44MSAxNyA1IDE5IDVhMSAxIDAgMCAxIDEgMXoiLz48cGF0aCBkPSJtOSAxMiAyIDIgNC00Ii8%2BPC9zdmc%2BCg%3D%3D)](https://github.com/tanviet12/vbsec)
  136. ```markdown
  137. [![vbsec: security scanned · passed](https://img.shields.io/badge/vbsec-security%20scanned%20%C2%B7%20passed-2ea44f?logo=data%3Aimage%2Fsvg%2Bxml%3Bbase64%2CPHN2ZyB4bWxucz0iaHR0cDovL3d3dy53My5vcmcvMjAwMC9zdmciIHZpZXdCb3g9IjAgMCAyNCAyNCIgZmlsbD0ibm9uZSIgc3Ryb2tlPSIjZmZmIiBzdHJva2Utd2lkdGg9IjIuMiIgc3Ryb2tlLWxpbmVjYXA9InJvdW5kIiBzdHJva2UtbGluZWpvaW49InJvdW5kIj48cGF0aCBkPSJNMjAgMTNjMCA1LTMuNSA3LjUtNy42NiA4Ljk1YTEgMSAwIDAgMS0uNjctLjAxQzcuNSAyMC41IDQgMTggNCAxM1Y2YTEgMSAwIDAgMSAxLTFjMiAwIDQuNS0xLjIgNi4yNC0yLjcyYTEuMTcgMS4xNyAwIDAgMSAxLjUyIDBDMTQuNTEgMy44MSAxNyA1IDE5IDVhMSAxIDAgMCAxIDEgMXoiLz48cGF0aCBkPSJtOSAxMiAyIDIgNC00Ii8%2BPC9zdmc%2BCg%3D%3D)](https://github.com/tanviet12/vbsec)
  138. ```
  139. ## Sponsors
  140. vbsec is free and open source thanks to:
  141. <table>
  142. <tr>
  143. <td align="center" valign="top" width="50%">
  144. <a href="https://sepay.vn?utm_source=github&utm_medium=readme&utm_campaign=vbsec"><img src="docs/images/sponsor-sepay.svg" alt="SePay" height="42"></a><br>
  145. <b><a href="https://sepay.vn?utm_source=github&utm_medium=readme&utm_campaign=vbsec">SePay</a></b><br>
  146. Open Banking platform: automatic bank transfer confirmation and API connections to Vietnamese banks
  147. </td>
  148. <td align="center" valign="top" width="50%">
  149. <a href="https://123host.vn?utm_source=github&utm_medium=readme&utm_campaign=vbsec"><img src="docs/images/sponsor-123host.svg" alt="123HOST" height="42"></a><br>
  150. <b><a href="https://123host.vn?utm_source=github&utm_medium=readme&utm_campaign=vbsec">123HOST</a></b><br>
  151. Hosting, VPS, servers and domains for businesses and developers in Vietnam
  152. </td>
  153. </tr>
  154. </table>
  155. ## Authors
  156. - **[Bui Tan Viet](https://www.facebook.com/buitanviet)** — CEO, [SePay](https://sepay.vn) and [123HOST](https://123host.vn)
  157. - **Phan Quoc Hien** — CTO, [SePay](https://sepay.vn) and [123HOST](https://123host.vn)
  158. vbsec distills security experience from running real payment and hosting systems, which get attacked every day, into a rule set that lets AI review the code AI writes.
  159. Other open-source projects:
  160. - **[Sano](https://github.com/tanviet12/sano-sach-noi)** — turn Word documents into audiobooks with AI; Vietnamese voices that run on your own machine
  161. - **[Chat Quality Agent](https://github.com/tanviet12/chat-quality-agent)** — AI scoring of customer-support quality on Zalo OA and Facebook Messenger
  162. ## Contributing
  163. Contributions are welcome: bug reports, rule fixes, new languages. See [docs/en/contributing.md](docs/en/contributing.md).
  164. ### Three platforms, one rule set
  165. | Platform | Skill folder | Large repos |
  166. |---|---|---|
  167. | Claude Code | `skills/vbs-scan-security/` (canonical) | 3 parallel agents |
  168. | OpenAI Codex CLI | `skills/codex/vbs-scan-security/` | Scans parts one by one |
  169. | Google Antigravity | `skills/antigravity/vbs-scan-security/` | Scans parts one by one |
  170. Edit rules in the canonical `skills/vbs-scan-security/`, then run `./scripts/sync-skills.sh` to copy them to the other two. `SKILL.md` and `workflows/large-review*.md` are maintained separately per platform.
  171. ### Before opening a PR
  172. ```bash
  173. ./scripts/sync-skills.sh # sync the 3 skill variants
  174. ./scripts/run-fixtures.sh typescript # scan the sample app and score it (costs tokens; run the language you changed)
  175. python3 scripts/check-lessons.py # every security lesson in your lessons file must map to a test case
  176. ```
  177. ### Roadmap
  178. - Done: 30 general rules, including 9 for SaaS and AI agents (v0.8); dedicated rules for Go, PHP, TypeScript/JavaScript, Python, .NET/C#, Dart/Flutter; three platforms; scanning without git; online CVE lookup via OSV.dev (`--sca`, opt-in, needs network); automatic fixes verified by building (`--auto-fix`, opt-in, needs git). See [usage](docs/en/usage.md#live-dependency-scan---sca)
  179. - Next: Ruby, Java, Rust, driven by community demand
  180. ## License
  181. [MIT](./LICENSE) · © 2026 Bui Tan Viet & Phan Quoc Hien

Ratings and comments

Sign in to rate this skill.

Sign in

Free to use. If it’s helped, pass it on.