π Ironwall β 7-Step Security Audit CLI

Open-source security audit CLI. 7-step pipeline. AI-assisted analysis. Your code never leaves your machine.
β οΈ Phase 1 (v0.1.0): Step 1 (gitleaks) is functional. Steps 2-7 are under active development.
π― What It Does
Ironwall scans your codebase through a 7-step security audit pipeline:
| Step |
Name |
Tool |
What It Finds |
| 1 |
π Secret Scanning |
gitleaks |
API keys, tokens, passwords in code |
| 2 |
π¬ SAST Analysis |
semgrep + AI |
SQL injection, XSS, command injection |
| 3 |
π Endpoint Audit |
AI |
Auth bypass, IDOR, missing access control |
| 4 |
π Hardcoded Secrets |
AI |
Patterns gitleaks missed |
| 5 |
π¦ Dependency CVE |
go/npm/pip |
Known vulnerabilities in dependencies |
| 6 |
π₯οΈ Server Config |
AI |
Nginx, Docker, env misconfigurations |
| 7 |
ποΈ Database Audit |
AI |
Migration risks, SQL anti-patterns |
Key principle: All scanning happens locally. AI analysis optionally sends code snippets (not your entire repo) to the API using your own key.
π¦ Installation
go install github.com/FYFran/ironwall/cmd/ironwall@latest
Or build from source:
git clone https://github.com/FYFran/ironwall.git
cd ironwall
make build
Requirements:
- Go 1.22+
- gitleaks (
go install github.com/gitleaks/gitleaks/v8@latest)
π Quick Start
# Scan current directory (terminal output)
ironwall scan .
# Quick scan β only secrets + hardcoded patterns (< 30s)
ironwall quick .
# Generate markdown report
ironwall scan . --format markdown
# Generate JSON report (for CI pipelines)
ironwall scan . --format json --output report.json
# Enable AI-assisted analysis
export DEEPSEEK_API_KEY="sk-..."
ironwall scan . --ai
π Example Output
π ironwall v0.1.0 β 7-Step Security Audit
ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
Target: ./my-app
Duration: 2.3s
π Secret scanning (gitleaks) .................... 2 found
π¬ SAST (semgrep + AI) ............................ 5 found
...
ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
π SUMMARY
π΄ CRITICAL: 1 π HIGH: 5 π‘ MEDIUM: 5 π’ LOW: 3
π Full report: ./ironwall-report-my-app.md
ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
π§ Configuration
| Flag |
Default |
Description |
--format |
terminal |
Output format: terminal, markdown, json |
--output, -o |
auto |
Output file path |
--quick |
false |
Only run fast steps (1+4) |
--ai |
false |
Enable AI-assisted analysis |
--ai-model |
deepseek-chat |
Model name (also supports OpenAI, Claude) |
--timeout |
300 |
Max scan time in seconds |
--verbose, -v |
false |
Verbose output |
Environment variables:
IRONWALL_AI_KEY or DEEPSEEK_API_KEY β Your AI API key
IRONWALL_AI_ENDPOINT β Custom API endpoint (default: DeepSeek)
ποΈ Architecture
ironwall CLI
βββ Pipeline Engine (sequential step execution)
βββ AI Engine (DeepSeek API β optional)
βββ Reporter Engine (terminal / markdown / JSON)
βββ External Tools (gitleaks, semgrep, nuclei, ...)
All external tools run as subprocesses on the user's machine. The AI engine uses the OpenAI-compatible API interface β works with DeepSeek, OpenAI, Claude, or local Ollama.
π Methodology
Ironwall follows a 7-step gated pipeline inspired by professional security audit workflows:
- Gated execution β Step 1 (gitleaks) is TIER1. If it fails, the scan aborts.
- Attack scenario verification β Every AI-generated finding must pass three questions:
- Q1: What role/conditions does the attacker need?
- Q2: What is the concrete attack path?
- Q3: What does the attacker gain?
- If all three have specific, concrete answers β real vulnerability. Otherwise β filtered.
- Gotchas library β Curated patterns that static analyzers typically miss.
Read the full methodology: docs/methodology.md
π Comparison
|
strix |
PentestMate |
Ironwall |
| Mode |
Open-source CLI |
Closed SaaS |
Open-source CLI |
| Price |
Free |
$59/mo |
Free (MIT) |
| Method |
Multi-agent pentest |
Scan-focused |
7-step gated + gotchas |
| False positives |
PoC-verified |
Low |
Three-question attack verification |
πΊοΈ Roadmap
- v0.1.0 β Step 1: Secret scanning (gitleaks)
- v0.2.0 β Steps 2+4: SAST + Hardcoded secrets with AI (545x faster)
- v0.3.0 β Steps 3+5+6+7: Endpoint audit + Dependencies + Server + Database
- v0.3.1 β Nuclei scanner, review command (diff-only), gitleaks tests, Fiverr gig
- v0.4.0 β 8-step pipeline: Betterleaks (98.6% recall), CodeQL, KICS, Syft/Grype, supply chain, AI engine (dual-model V3+R1), SARIF, CI templates
- v0.5.0 β Custom rule engine, plugin system, language-specific gotchas expansion
- v1.0.0 β Stable API, comprehensive benchmarks, marketplace integration
π€ Contributing
Contributions welcome! Especially:
- Gotchas β Patterns your tools missed. Add to
docs/gotchas.md.
- Test data β Vulnerable code samples for
testdata/.
- Language support β Scanner modules for new languages.
See CONTRIBUTING.md (coming soon).
π License
MIT Β© 2026 FYFran
Built by @FYFran β CS freshman at Taizhou University. Learning security by building tools.