How cyberstrikeus/cyberstrike works
CyberStrike is a terminal-first AI agent for offensive security testing. The principal workflow is an operator request, agent-led testing with an LLM and security skills/tools, then findings and a report. The tree also shows a web app, HTTP server, MCP and LSP commands, and persistent sessions. The README documents broader capabilities, including Bolt remote execution and many specialized agents; sampled source is incomplete, so uncertain implementation wiring is omitted.
Main entry points: agent.ts Agent runtime · cmd.ts CLI commands · app.tsx Terminal UI · index.ts LLM providers
Components 20
User Interfaces
- CLI commands
packages/cyberstrike/src/cli/cmd/cmd.ts - Terminal UI
packages/cyberstrike/src/cli/cmd/tui/app.tsx - Web app
packages/app/src/app.tsx
Agent Intelligence
- Agent runtime
packages/cyberstrike/src/agent/agent.ts - LLM providers
packages/cyberstrike/src/auth/index.ts - Skill index
packages/cyberstrike/src/skill/index-engine.ts - Specialized agents
packages/cyberstrike/src/agent/agent.ts
Security Testing
- Reconnaissance
packages/cyberstrike/src/cli/cmd/run.ts - Vulnerability discovery
packages/cyberstrike/src/cli/cmd/run.ts - Exploitation
packages/cyberstrike/src/cli/cmd/run.ts - Findings and reports
packages/cyberstrike/src/cli/cmd/tui/component/dialog-report.tsx
Runtime Services
- MCP commands
packages/cyberstrike/src/cli/cmd/mcp.ts - Language servers
packages/cyberstrike/src/lsp/server.ts - HTTP server
packages/cyberstrike/src/server/server.ts - Session storage
packages/cyberstrike/src/storage/storage.ts - Browser capture
packages/hackbrowser/src/ingest.ts - Bolt remote execution
packages/cyberstrike/src
Other components
- Security operator
- Authorized targets
- LLM service
Connections 20
- Security operator → Terminal UI: submits test
- CLI commands → Terminal UI: launches
- Terminal UI → Agent runtime: requests testing
- Web app → HTTP server: connects to
- Agent runtime → LLM providers: uses model
- LLM providers → LLM service: sends prompts
- Agent runtime → Skill index: retrieves skills
- Agent runtime → Specialized agents: selects agents
- Agent runtime → Reconnaissance: orchestrates
- Agent runtime → Vulnerability discovery: orchestrates
- Agent runtime → Exploitation: orchestrates
- Reconnaissance → Authorized targets: tests
- Vulnerability discovery → Authorized targets: tests
- Exploitation → Authorized targets: tests
- Agent runtime → Findings and reports: produces findings
- Agent runtime → Session storage: persists sessions
- CLI commands → MCP commands: dispatches
- CLI commands → Language servers: dispatches
- Browser capture → HTTP server: submits capture
- Agent runtime → Bolt remote execution: uses remote execution