Skip to content

camgrimsec/grimsec-suite

Repository files navigation

GRIMSEC — AI-Powered DevSecOps Agent Suite

12 autonomous security agents that analyze, audit, validate, and remediate vulnerabilities in any GitHub repository. Context-aware. Noise-reduced. PR-ready.


Why GRIMSEC?

Traditional scanners dump hundreds of CVEs. GRIMSEC's 12-agent pipeline tells you which ones actually matter.

The problem: Run Trivy on any repo → get 500+ findings. How many are actually exploitable? Maybe 5.

GRIMSEC's answer: 6-stage reachability analysis → Real Risk Scores → exploit validation → auto-generated PRs → executive reports. 89-96% noise reduction.


Quick Start

git clone https://github.com/yourusername/grimsec-suite.git
cd grimsec-suite
bash setup.sh
python grimsec.py analyze https://github.com/org/repo

See docs/quickstart.md for a detailed 5-minute guide.


The 12 Agents

# Agent What It Does
1 Repo Analyzer 6-stage pipeline: inventory → STRIDE → scan → reachability → remediate → report
2 CI/CD Auditor Supply chain security: unpinned actions, expression injection, PPE, secrets
3 Vuln Enricher NVD + EPSS + CISA KEV + ATT&CK mapping per CVE
4 Doc Intelligence Reads project docs to validate/downgrade scanner findings
5 Threat Intel Monitor Continuous CVE monitoring against your dependency inventory
6 Executive Reporter Translates findings into $$ risk, compliance mapping, board-ready reports
7 DAST Scanner Nuclei + ZAP against running applications
8 Exploit Validator Generates PoCs proving findings are actually exploitable
9 Code Understanding Attack surface mapping, source-to-sink tracing, variant hunting
10 IaC Policy Checkov + OPA: Docker, K8s, Terraform, GitHub Actions policies
11 OSS Forensics Supply chain investigation: evidence collection, IOC detection, timeline
12 Adversary Sim Controlled exploitation + ATT&CK mapping

Architecture

┌─────────────────────────────────────────────────────────────────┐
│                    GRIMSEC PIPELINE                             │
│                                                                 │
│  INPUT: GitHub Repo URL                                         │
│         │                                                       │
│         ▼                                                       │
│  ┌─────────────┐    ┌─────────────┐    ┌─────────────────────┐ │
│  │  Agent 01   │    │  Agent 02   │    │     Agent 03        │ │
│  │  Repo       │───▶│  CI/CD      │───▶│  Vuln Enricher      │ │
│  │  Analyzer   │    │  Auditor    │    │  (NVD+EPSS+KEV)     │ │
│  └─────────────┘    └─────────────┘    └─────────────────────┘ │
│         │                  │                      │             │
│         ▼                  ▼                      ▼             │
│  inventory.json     cicd-findings.json    enriched-cvelist.json │
│         │                                         │             │
│         ▼                                         ▼             │
│  ┌─────────────┐                        ┌─────────────────────┐ │
│  │  Agent 04   │                        │     Agent 05        │ │
│  │  Doc Intel  │                        │  Threat Intel       │ │
│  │  (context)  │                        │  Monitor            │ │
│  └─────────────┘                        └─────────────────────┘ │
│         │                                         │             │
│         └───────────────────┬───────────────────-─┘             │
│                             ▼                                   │
│              ┌──────────────────────────┐                       │
│              │  Agent 06: Exec Reporter │                       │
│              │  • $$ Risk quantification│                       │
│              │  • Compliance mapping    │                       │
│              │  • Board-ready deck      │                       │
│              └──────────────────────────┘                       │
│                                                                 │
│  ── DEEP MODE adds ─────────────────────────────────────────── │
│                                                                 │
│  ┌────────────┐  ┌────────────┐  ┌────────────┐  ┌──────────┐ │
│  │ Agent 07   │  │ Agent 08   │  │ Agent 09   │  │Agent 10  │ │
│  │ DAST       │  │ Exploit    │  │ Code       │  │ IaC      │ │
│  │ Scanner    │  │ Validator  │  │ Understanding│ │ Policy   │ │
│  └────────────┘  └────────────┘  └────────────┘  └──────────┘ │
│                                                                 │
│  ┌────────────┐  ┌────────────┐                                │
│  │ Agent 11   │  │ Agent 12   │                                │
│  │ OSS        │  │ Adversary  │                                │
│  │ Forensics  │  │ Simulation │                                │
│  └────────────┘  └────────────┘                                │
└─────────────────────────────────────────────────────────────────┘

See docs/architecture.md for detailed agent chaining and data flow.


Usage with Perplexity Computer

Each agent is a Perplexity Computer skill. To import them:

  1. Go to perplexity.ai → Computer → Skills
  2. Upload the SKILL.md from the agent directory (or zip the whole agent folder)
  3. Trigger with natural language:
    • "analyze this repo for security issues: https://github.com/org/repo"
    • "audit the CI/CD pipeline at this URL"
    • "enrich CVE-2024-1234"
    • "generate an executive security report"

Skill Loading Reference

Command Skill to Load Trigger Phrase
analyze devsecops-repo-analyzer "analyze this repo for security issues"
audit cicd-pipeline-auditor "audit the CI/CD pipeline"
enrich vulnerability-context-enricher "enrich CVE-XXXX-XXXXX"
doc doc-intelligence-agent "read the project documentation"
monitor threat-intel-monitor "run threat intel check"
report executive-reporting-agent "generate executive report"
dast dast-scanner "run DAST scan against this URL"
validate exploit-validation-agent "validate these findings"
understand code-understanding-agent "map the attack surface"
iac iac-policy-agent "run IaC policy scan"
forensics oss-forensics-agent "investigate this package for supply chain compromise"
simulate adversary-simulation-agent "run adversary simulation"

CLI Usage

# Full 12-agent pipeline
python grimsec.py analyze https://github.com/org/repo

# Quick scan (Agents 1-3 only, ~10 min)
python grimsec.py analyze https://github.com/org/repo --quick

# Deep scan (all agents + DAST + adversary sim)
python grimsec.py analyze https://github.com/org/repo --deep

# Individual agent commands
python grimsec.py scan https://github.com/org/repo      # Agent 1 (SCA+SAST+secrets)
python grimsec.py audit https://github.com/org/repo     # Agent 2 (CI/CD audit)
python grimsec.py enrich CVE-2024-1234                  # Agent 3 (CVE enrichment)
python grimsec.py monitor                               # Agent 5 (threat intel)
python grimsec.py report grimsec-output/repo/2026-01-01 # Agent 6 (exec report)
python grimsec.py dast https://staging.example.com      # Agent 7 (DAST)
python grimsec.py validate grimsec-output/repo/2026-01-01 # Agent 8 (PoC validation)
python grimsec.py understand https://github.com/org/repo # Agent 9 (attack surface)
python grimsec.py iac /path/to/repo                     # Agent 10 (IaC policy)
python grimsec.py forensics https://github.com/org/repo # Agent 11 (supply chain)
python grimsec.py simulate staging.example.com          # Agent 12 (adversary sim)

# Utilities
python grimsec.py status    # Check all tools installed
python grimsec.py install   # Install all tools
python grimsec.py dashboard # Dashboard setup instructions

Output Structure

After a run, findings are organized at grimsec-output/<repo-name>/<timestamp>/:

grimsec-output/example-repo/2026-01-01T12-00-00/
├── inventory.json                # Dependencies, tech stack, entry points
├── app-context.json              # Application type, architecture, threat model
├── stride-threats.json           # STRIDE threat enumeration
├── scan-results/
│   ├── trivy-sca.json            # SCA findings
│   ├── semgrep-sast.json         # SAST findings
│   ├── gitleaks-secrets.json     # Secrets scan
│   └── grype-sbom.json           # SBOM + SCA
├── reachability-analysis.json    # Which findings are actually reachable
├── enriched-findings.json        # CVEs enriched with EPSS + KEV + ATT&CK
├── cicd-findings.json            # CI/CD audit results
├── doc-context.json              # Documentation intelligence
├── iac-findings.json             # IaC policy violations
├── dast-results/                 # DAST scan output
├── exploit-validation.json       # PoC-backed validation results
├── attack-surface-map.json       # Attack surface + data flow
├── forensics-report.json         # Supply chain forensics
├── simulation-report.json        # Adversary simulation results
└── executive-summary.json        # Board-ready executive report

See examples/sample-output/README.md for full structure.


Proven Results

Metric Value
Average noise reduction 89-96%
Finding validation method PoC-backed
Compliance frameworks covered SOC 2, ISO 27001, NIST CSF, OWASP SAMM
CVE enrichment sources NVD, EPSS, CISA KEV, MITRE ATT&CK
Supported languages Python, JavaScript/TypeScript, Go, Java, Ruby, Rust
IaC targets Dockerfile, Kubernetes, Terraform, GitHub Actions

Requirements

Tool Purpose Required?
Python 3.8+ CLI runner Yes
Trivy SCA + container scanning Yes
Semgrep SAST Yes
Gitleaks Secrets detection Yes
Grype SCA (SBOM-based) Yes
Nuclei DAST (Agent 7) For DAST
httpx HTTP probing (Agent 7) For DAST
Checkov IaC scanning (Agent 10) For IaC
OPA Policy engine (Agent 10) For IaC
Conftest Config testing (Agent 10) For IaC
Syft SBOM generation (Agent 10) For IaC
Snyk CLI Additional SCA Optional

Run bash setup.sh to install everything automatically.


Contributing

Contributions are welcome. See CONTRIBUTING.md for guidelines.

Areas where help is most wanted:

  • New Semgrep rules for additional languages
  • Additional Nuclei templates for DAST coverage
  • New OPA policies for cloud providers
  • Dashboard integrations (Grafana, Metabase, Kibana)
  • Integrations with JIRA, GitHub Issues, PagerDuty

Security

See SECURITY.md for the security policy and how to report vulnerabilities.


License

MIT — see LICENSE


Legal

GRIMSEC is for authorized security testing only. Only run against systems you own or have explicit written permission to test. The exploit validation and adversary simulation agents include safeguards, but you are responsible for ensuring all usage is authorized and legal in your jurisdiction.

About

GRIMSEC — 12 AI-Powered DevSecOps Agents. Analyze, audit, validate, and remediate vulnerabilities in any GitHub repo. 89-96% noise reduction.

Topics

Resources

License

Contributing

Security policy

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors