bmad-hardened
v6.2.0
Published
Security-hardened fork of BMad Method — Agile AI-driven Development with integrated security workflows
Maintainers
Readme
BMAD-HARDENED
Community fork of BMad Method with integrated security hardening across the entire agile lifecycle.
Credit: This project is a fork of BMad Method by Brian (BMad) Madison / BMad Code, LLC. The entire base framework, agent architecture, workflows, and CLI are his work. BMAD-HARDENED adds a security layer on top of this foundation.
What BMAD-HARDENED adds
3 new security agents (Party Mode)
| Agent | Name | Role | |-------|------|------| | 🛡️ Cybersecurity Expert | Nyx | Vulnerability analysis, OWASP, CVE tracking, supply chain, LLM security | | 🏰 Security Architect | Bastion | Threat modeling STRIDE/DREAD, zero-trust, isolation, crypto design | | 🤓 Tech Genius | Zero | Bleeding-edge tech watch, alternatives to mainstream solutions |
1 new workflow: Security Review
Comprehensive security audit of architecture, PRD, and stories before implementation. Integrated in Phase 3 (Solutioning).
24 universal security DATA files
Security knowledge base loaded on demand by any agent/workflow via a tag-based index system:
- 11 attack patterns (
atk-*): LLM injection, supply chain, privilege escalation, reverse engineering... - 10 defense patterns (
def-*): crypto, auth, OS isolation, framework hardening, audit logging... - 3 reference files (
ref-*): agent threat model, cross-validation matrix, CVE catalog
All files are language/framework agnostic — they describe universal patterns, not specific implementations.
Reinforcement of existing agents and workflows
- All agents: systematic web search before any work + global rules compliance
- Code Review: adversarial security deep dive added to workflow
- 5 workflows enriched with conditional security data loading (INDEX_THEN_SELECTIVE)
- Global Agent Rules: mandatory web search, feature checkup tracking, story size enforcement, review scope guard
- Implementation Readiness: blocking security gate before Phase 4
Loading strategy: INDEX_THEN_SELECTIVE
Security data is never loaded in bulk. The mechanism:
- Agent/workflow loads
index.md(file listing with tags) - Matches tags against current context (stack, domain, story)
- Loads only 3-5 relevant files
3 defense layers ensure loading:
- Global rules reference index.md
- Each workflow has explicit loading instructions
- Agent critical_actions reference index.md
Installation
npx bmad-hardened installInstalls the full BMAD-HARDENED framework (agents, workflows, security data, templates) into your project.
Original project
BMad Method by Brian (BMad) Madison / BMad Code, LLC
Changelog
See FORK-CHANGES.md for the complete list of modifications made by this fork.
See CHANGELOG.md for the original project history.
License
MIT License — see LICENSE.
BMad and BMAD-METHOD are trademarks of BMad Code, LLC. See TRADEMARK.md. This fork is an unofficial community project, not affiliated with or endorsed by BMad Code, LLC.
