<p align="center">
<img src="assets/logo.svg" width="80" alt="foxguard logo" />
</p>
<h1 align="center">foxguard</h1>
<p align="center">
<strong>Sub-second local security scanning for real codebases.</strong>
<br/>
130+ built-in rules · 10 languages · taint tracking for Python, JavaScript, Go · single Rust binary · Semgrep-compatible YAML bridge
<br/><br/>
<a href="https://foxguard.dev">foxguard.dev</a> · <a href="https://www.npmjs.com/package/foxguard">npm</a> · <a href="https://crates.io/crates/foxguard">crates.io</a>
</p>
<p align="center">
<a href="https://github.com/PwnKit-Labs/foxguard/actions/workflows/ci.yml"><img src="https://github.com/PwnKit-Labs/foxguard/actions/workflows/ci.yml/badge.svg" alt="CI" /></a>
<a href="https://github.com/PwnKit-Labs/foxguard"><img src="https://img.shields.io/badge/foxguard-clean-2dd4bf?logo=data:image/svg%2bxml;base64,PHN2ZyB4bWxucz0iaHR0cDovL3d3dy53My5vcmcvMjAwMC9zdmciIHZpZXdCb3g9IjAgMCA2NCA2NCIgZmlsbD0ibm9uZSI+PHBhdGggZD0iTTggOEwyMCAyOEwzMiAyMEw0NCAyOEw1NiA4TDUyIDMyTDQ0IDQ0TDM2IDUySDI4TDIwIDQ0TDEyIDMyTDggOFoiIGZpbGw9IiNGNTlFMEIiIGZpbGwtb3BhY2l0eT0iMC4zIiBzdHJva2U9IiNGNTlFMEIiIHN0cm9rZS13aWR0aD0iMyIgc3Ryb2tlLWxpbmVqb2luPSJyb3VuZCIvPjxjaXJjbGUgY3g9IjI0IiBjeT0iMzIiIHI9IjIuNSIgZmlsbD0iI0Y1OUUwQiIvPjxjaXJjbGUgY3g9IjQwIiBjeT0iMzIiIHI9IjIuNSIgZmlsbD0iI0Y1OUUwQiIvPjwvc3ZnPg==" alt="foxguard: clean" /></a>
<a href="https://crates.io/crates/foxguard"><img src="https://img.shields.io/crates/v/foxguard?color=d97706&label=crates.io" alt="crates.io" /></a>
<a href="https://www.npmjs.com/package/foxguard"><img src="https://img.shields.io/npm/v/foxguard?color=d97706&label=npm" alt="npm" /></a>
</p>
---
<p align="center">
<img src="assets/demo.gif" alt="foxguard vs semgrep side-by-side" width="640" />
</p>
Security scanners are slow. 10 seconds, 30 seconds, sometimes a minute. So developers don't run them locally — they get pushed to CI, findings pile up in PRs, and nobody looks at them.
foxguard fixes this by being fast enough that you never notice it's there. Same scan, 0.03 seconds instead of 10. You can run it on every save, every commit, every push. Security feedback becomes instant.
```sh
npx foxguard .
```
```
src/auth/login.js
14:5 CRITICAL js/no-sql-injection (CWE-89)
SQL query built with template literal interpolation
src/utils/config.py
7:1 HIGH py/no-hardcoded-secret (CWE-798)
Hardcoded secret in 'api_key'
WARNING 2 issues in 5 files (0.03s): 1 critical, 1 high, 0 medium, 0 low
```
## Why foxguard
- **Fast enough to leave on.** foxguard is built for local runs, pre-commit hooks, and changed-file scans instead of “security later in CI”.
- **Useful before you tune anything.** The default value is built-in framework-aware rules for common real-world mistakes across JavaScript, Python, Go, Ruby, Java, PHP, Rust, C#, and Swift.
- **Taint tracking built in.** Intraprocedural taint flow from framework sources (Flask, Django, FastAPI, Express, Next.js, Hono, Gin, net/http) into sinks like `eval`, `exec`, SQL execute, and SSRF — no rule writing required.
- **Adoption-friendly.** If you already have Semgrep/OpenGrep YAML, foxguard can load a focused compatible subset on top of built-ins so migration is incremental instead of all-or-nothing.
See [docs/precision.md](docs/precision.md) for per-rule precision tiers and our false-positive methodology.
## Quick start
```sh
npx foxguard . # scan the repo
npx foxguard --changed . # only modified files
npx foxguard secrets . # leaked credentials and private keys
npx foxguard init # install a local pre-commit hook
```
## What it is
Rust + [tree-sitter](https://tree-sitter.github.io/) for AST parsing + [rayon](https://github.com/rayon-rs/rayon) for parallelism. No JVM startup, no Python interpreter, no network calls, no rule download step. Just a native binary that reads your files and reports findings.
130+ built-in rules across 10 languages. SQL injection, XSS, SSRF, command injection, hardcoded secrets, weak crypto, unsafe deserialization, log injection, and framework-specific checks for Express, Django, Rails, Spring, Laravel, Gin, .NET, and iOS. Python, JavaScript, and Go also get an intraprocedural taint engine that follows untrusted input from framework request sources into dangerous sinks.
Also scans for leaked credentials (AWS keys, GitHub/GitLab/Slack/Stripe tokens, private keys) with redacted output. Loads Semgrep-compatible YAML rules with `--rules` if you have existing ones. Outputs terminal, JSON, or SARIF for GitHub Code Scanning.
foxguard dogfoods itself — it scans its own Rust source in CI on every push.
## What it is not
foxguard is not trying to be a full Semgrep or OpenGrep drop-in replacement.
The intended model is:
- **foxguard built-ins** for fast local feedback
- **Semgrep/OpenGrep-compatible YAML subset** as an adoption bridge
- **Semgrep/OpenGrep themselves** when you need the broadest external rule ecosystem
That boundary is deliberate. It keeps local scans fast, rule support understandable, and compatibility claims testable.
## Install
```sh
npx foxguard . # no install needed
brew install peaktwilight/tap/foxguard # Homebrew (macOS/Linux)
cargo install foxguard # crates.io
```
**Editor:** Install the [VS Code extension](https://marketplace.visualstudio.com/items?itemName=peaktwilight.foxguard) — scans on save, shows findings as underlines.
## Benchmarks
Reproducible benchmarks via `./benchmarks/run.sh`. Numbers below are from a local run on an Apple Silicon laptop with `foxguard 0.5.0`, `semgrep 1.156.0`, `tokei 14.0.0`. LoC is counted by tokei, scoped to the target language only (no vendored HTML/JSON).
| Repo | Files | LoC | foxguard | Semgrep | Speedup |
|------|-------|-----|----------|---------|---------|
| express (framework) | 141 | 15,804 JS | **0.11s** | 4.80s | **45x** |
| flask (framework) | 83 | 14,029 Py | **0.08s** | 5.70s | **73x** |
| gin (framework) | 99 | 17,669 Go | **0.07s** | 4.61s | **63x** |
| **sentry (production)** | **8,539** | **1,291,606 Py** | **12.19s** | 164.53s | **13x** |
Sentry is the larger-corpus stress target added under issue #8: a real production monitoring platform at ~1.3M Python LoC. foxguard scans the whole tree in ~12 seconds (~106k LoC/sec); Semgrep with `--config auto` takes ~2m45s on the same tree. Run on one machine — your numbers will vary; reproduce locally with `./benchmarks/run.sh`.
To reproduce: `./benchmarks/run.sh` (add `BENCH_SKIP_LARGE=1` for the quick matrix only). See `benchmarks/README.md` for the reproduction recipe.
## Built-in coverage
| Language | Rules | Frameworks |
|----------|-------|------------|
| JavaScript/TypeScript | 27 | Express, Next.js, Hono, Fastify, SvelteKit, Deno, JWT, XSS, taint |
| Python | 32 | Flask, Django, FastAPI, CSRF, session, intraprocedural taint |
| Go | 11 | Gin, net/http, TLS, intraprocedural taint |
| Ruby | 10 | Rails, mass assignment, CSRF |
| Java | 10 | Spring, XXE, deserialization |
| PHP | 10 | Laravel, file inclusion, unserialize |
| Rust | 10 | unsafe, transmute, TLS |
| C# | 10 | .NET, LDAP, XXE, CORS |
| Swift | 10 | iOS keychain, transport, WebView |
## Why teams adopt it
- **Changed-file scans** for tight local loops
- **Repo-local baselines** so legacy findings stop blocking adoption
- **Secrets scanning** alongside code scanning
- **JSON and SARIF output** for CI and GitHub Code Scanning
- **Semgrep/OpenGrep YAML subset** when teams already have rule investments
## Compatibility
Load existing Semgrep/OpenGrep YAML rules with `--rules`. Supports `pattern`, `pattern-regex`, `pattern-either`, `pattern-not`, `pattern-inside`, `pattern-not-inside`, `metavariable-regex`, and `paths.include/exclude`. This supported subset is parity-tested in CI against the real `semgrep` CLI. See [`COMPATIBILITY.md`](./COMPATIBILITY.md).
foxguard does not currently aim to support multiple unrelated external rule formats. The compatibility target is the focused Semgrep/OpenGrep YAML subset above.
## CI Integration
### GitHub Actions
```yaml
name: Security
on: [push, pull_request]
jobs:
foxguard:
runs-on: ubuntu-latest
permissions:
security-events: write
steps:
- uses: actions/checkout@v4
- uses: PwnKit-Labs/foxguard/action@v0.5.0
with:
path: .
severity: medium
fail-on-findings: "true"
upload-sarif: "true"
```
Findings show up in **Security → Code Scanning**.
### Any CI
```sh
npx foxguard@latest . # scan
npx foxguard@latest --format sarif . > out.sarif # SARIF output
npx foxguard@latest secrets . # secrets
```
### Badge
```md
[](https://github.com/PwnKit-Labs/foxguard)
```
### Pre-commit
```yaml
repos:
- repo: https://github.com/PwnKit-Labs/foxguard
rev: v0.5.0
hooks:
- id: foxguard
- id: foxguard-secrets
```
Or run `foxguard init` to install a git hook directly.
### Claude Code
Add foxguard as a pre-commit hook in your Claude Code configuration to automatically scan agent-written code before each commit:
```json
{
"hooks": {
"PreCommit": [
{
"command": "npx foxguard --changed --severity high .",
"description": "foxguard security scan"
}
]
}
}
```
Add this to `.claude/settings.json` in your project root. Claude Code will run foxguard before every commit and block if high-severity findings are detected, giving the agent a chance to fix issues before they land. See [docs/claude-code-integration.md](docs/claude-code-integration.md) for the full setup guide.
## Configuration
foxguard auto-discovers `.foxguard.yml` from the scan path upward.
```yaml
scan:
baseline: .foxguard/baseline.json
rules: ./semgrep-rules
secrets:
baseline: .foxguard/secrets-baseline.json
exclude_paths:
- fixtures
- testdata
ignore_rules:
- secret/github-token
```
## Suppressing Deliberate Findings
For one-off, deliberate code patterns, you can suppress code-scan findings inline instead of
adding them to a baseline.
```js
// Ignore the next code line for one rule
// foxguard: ignore[js/no-ssrf]
fileContent = fetch(userControlledUrl);
// Ignore the current line for one rule
fileContent = fetch(userControlledUrl); // foxguard: ignore[js/no-ssrf]
// Ignore the current line for all foxguard code findings
eval(userInput); // foxguard: ignore
```
Notes:
- Inline ignores currently apply to code scanning findings, not `foxguard secrets`.
- Rule IDs must match exactly, for example `js/no-ssrf`.
- Comment-only directives apply to the next non-empty, non-comment code line.
- Supported comment styles are `//` and `#`, depending on the language.
## Contributing
Adding a rule is one struct implementing a trait. See [`CONTRIBUTING.md`](./CONTRIBUTING.md).
## Part of PwnKit Labs
**Open-source adversarial security for the agentic AI era.** foxguard is one piece of the open-source PwnKit Labs stack:
- **[pwnkit](https://github.com/PwnKit-Labs/pwnkit)** — AI agent pentester (detect)
- **[foxguard](https://github.com/PwnKit-Labs/foxguard)** — Rust security scanner (prevent)
- **[opensoar](https://github.com/opensoar-hq/opensoar-core)** — Python-native SOAR platform (respond)
## License
MIT