T3MP3ST Feature Documentation
July 5, 2026 · View on GitHub
Comprehensive feature list for the Tactical Execution Multi-agent Platform for Elite Security Testing
Legend:
- Implemented and working
- [~] Partially implemented / stub
- Planned / ideal feature
Table of Contents
- Core Architecture
- LLM Backbone
- Operator System (Agents)
- Mission Control
- Target Environment
- Arsenal (Tools)
- Pliny Specials
- Evidence Vault
- OPSEC Layer
- Communications
- Analysis & Reporting
- CLI Interface
- API Server
- MCP Server (Agent Integration)
- Web UI Dashboard
- Advanced Modules
- Configuration System
- Integration & Extensibility
1. Core Architecture
TempestCommand Orchestrator
- Central command class extending EventEmitter
- Lifecycle management (start, stop, pause, resume)
- Tick-based execution loop
- Event emission for all major actions
- Status reporting and health checks
- [~] Hooks system for lifecycle events
- Distributed command across multiple nodes
- State persistence and recovery after crash
Factory Functions
-
createTempest()- Standard operation -
createTestTempest()- Testing with mock LLM -
createAutoTempest()- Auto-configured with best available LLM -
createStealthOperation()- Silent OPSEC preset -
createAggressiveOperation()- Loud OPSEC preset -
createDistributedTempest()- Multi-node operation -
createCloudTempest()- Cloud-native deployment
Event System
-
command:started/command:stopped -
command:paused/command:resumed -
operator:spawned/operator:burned -
finding:discovered -
credential:harvested -
target:owned -
detection:triggered -
mission:phase_changed -
abort:recommended -
chain:completed- Full kill chain success -
exfil:complete- Data exfiltration finished
2. LLM Backbone
Provider Support
- OpenRouter - Multi-model gateway (50+ models)
- Anthropic - Direct Claude API
- OpenAI - GPT models
- Mock - Testing without API
- Local - Ollama support
- Azure OpenAI - Enterprise deployment
- AWS Bedrock - Claude via AWS
- Google Vertex AI - Gemini via GCP
Model Registry (50+ models configured)
- Claude Opus 4.5, Sonnet 4.5, Sonnet 4
- GPT-4o, GPT-4 Turbo, o1
- Gemini 3 Pro, Gemini 3 Flash, Gemini 2.5 Flash
- Grok 4, Grok 4 Fast, Grok 4.1 Fast
- DeepSeek R1, DeepSeek V3
- Llama 3.3 70B
- Mistral Large
- GLM 4.7
LLM Features
- Chat completion with retry logic
- Conversation history management
- System prompt configuration
- Streaming responses (OpenRouter)
- Token usage tracking
- Provider validation
- Automatic provider selection
- Function calling / tool use
- Vision/multimodal support
- Embeddings for semantic search
- Fine-tuned model support
- Cost tracking and budgeting
3. Operator System (Agents)
8 Operator Archetypes
| Archetype | Phase | MITRE Tactics | Status |
|---|---|---|---|
| RECON | Reconnaissance | TA0043 | [x] Implemented |
| SCANNER | Discovery | TA0007 | [x] Implemented |
| EXPLOITER | Initial Access | TA0001, TA0002 | ⚠️ Experimental¹ |
| INFILTRATOR | Lateral Movement | TA0008, TA0004 | ⚠️ Experimental¹ |
| EXFILTRATOR | Exfiltration | TA0009, TA0010 | ⚠️ Experimental¹ |
| GHOST | Persistence | TA0003, TA0005 | ⚠️ Experimental¹ |
| COORDINATOR | C2 | TA0011 | ⚠️ Experimental¹ |
| ANALYST | Reporting | - | [x] Implemented |
¹ Experimental = runs the SAME real, tool-backed ReAct loop as RECON/SCANNER (real tools, not stubs), but end-to-end swarm exploitation is unbenchmarked and unproven — 0 executed exploits in full-chain runs, and the headline benchmark numbers came from a single agent, not the coordinated 8-operator cell. See the README status table + WHITEPAPER for the honest limits.
Operator Features
- State management (idle, tasked, executing, cooldown, burned)
- Task assignment and completion tracking
- Detection risk monitoring
- Cooldown management
- Finding and credential recording
- Archetype-specific system prompts
- MITRE ATT&CK technique mapping
- [~] Autonomous decision making via LLM
- Inter-operator coordination
- Learning from past engagements
- Skill leveling system
- Custom archetype creation
Team Factories
-
createBalancedTeam()- General purpose -
createStealthTeam()- Minimal detection -
createBreachTeam()- Aggressive initial access -
createAPTTeam()- Advanced persistent threat simulation -
createInsiderTeam()- Insider threat simulation -
createCloudTeam()- Cloud-focused operations
Operator Cell Management
- Spawn/despawn operators
- Status aggregation
- Count by status
- Automatic scaling based on workload
- Operator health monitoring
- Failover and replacement
4. Mission Control
Mission Lifecycle
- Create mission with objectives
- Start/pause/resume/complete/abort
- Rules of Engagement (RoE) enforcement
- Phase transitions
- [~] Objective tracking and completion
- Mission templates
- Multi-mission management
- Mission scheduling
Task Queue
- Priority-based task management
- Add/remove tasks
- Get next task (global or by archetype)
- Status updates
- Mission-scoped task queries
- Task dependencies
- Parallel task execution
- Task retry with backoff
- Deadline enforcement
Rules of Engagement
- Scope definition (in-scope targets)
- Forbidden techniques list
- Detection event limits
-
createDefaultRoE()preset -
createStrictRoE()preset - Time-based restrictions
- Geographic restrictions
- Data handling rules
- Escalation procedures
Kill Chain Phases
- RECON - Reconnaissance
- WEAPONIZE - Weaponization
- DELIVER - Delivery
- EXPLOIT - Exploitation
- INSTALL - Installation
- C2 - Command & Control
- ACTIONS_ON_OBJECTIVES - Actions on Objectives
- Phase-to-archetype mapping
- Custom phase definitions
- Phase-specific tooling
5. Target Environment
Target Management
- Add/update/remove targets
- Status transitions (discovered → scanning → vulnerable → exploited → owned)
- Zone classification (external, dmz, internal, restricted)
- Type classification (web_app, api, server, network_device, etc.)
- Service and vulnerability tracking
- Statistics aggregation
- Asset discovery automation
- Network topology mapping
- Relationship tracking between targets
Target Factories
-
createTargetFromUrl()- Parse URL into target -
createTargetFromIP()- Create from IP address -
createDMZArchitecture()- Sample architecture -
importFromNmap()- Import from Nmap XML -
importFromShodan()- Import from Shodan -
importFromCloud()- Import from AWS/GCP/Azure
Target Types
- web_application
- api
- server
- workstation
- network_device
- database
- cloud_service
- container
- iot_device
- mobile_app
6. Arsenal (Tools)
Built-in Tools
-
dns_lookup- DNS resolution with multiple record types -
port_scan- Port scanning with service detection -
http_request- Full HTTP client with headers/body -
subdomain_enum- Subdomain enumeration -
dir_bruteforce- Directory bruteforcing with wordlists -
whois_lookup- Domain WHOIS lookup -
header_analysis- Security header analyzer -
technology_detect- Tech stack fingerprinting -
xss_scan- XSS vulnerability testing -
sqli_scan- SQL injection testing -
ssl_scan- TLS/SSL configuration analysis -
password_spray- Common password testing -
hash_crack- Hash identification and cracking -
base64_decode- Base64 decoding utility -
jwt_decode- JWT analysis with security checks -
vuln_scan- Comprehensive vulnerability scanning -
exploit_search- Exploit database search
Tool Management
- Tool registration
- Category-based filtering
- Tool execution with context
- Execution history tracking
- Custom tool support
- Tool chaining
- Tool output parsing
- Tool version management
- Tool health checks
Whitelisted CLI Tools (via API)
Availability tiers: default = wired without extra flags; opt-in = catalog-gated behind
T3MP3ST_FULL_ARSENAL; approval-gated = also requires per-call approval (real auth/attack traffic). See the35 default / 83 opt-insplit in the README. Real, re-derivable catalog (verify-claims): 35 built-in (default) + 48 opt-in adapters = 83.
- default built-ins (callable with no flag) —
nmap_scan, nuclei_scan, ffuf_fuzz, curl_request, port_scan, subdomain_enum, http_request, xss_scan, sqli_scan, dns_lookup, … (35) - opt-in adapters (
T3MP3ST_FULL_ARSENAL) — nuclei, subfinder, httpx, naabu, katana, ffuf, gobuster, feroxbuster, nikto, dalfox, sqlmap · semgrep, gitleaks, trufflehog, trivy, grype, osv-scanner, checkov · slither, mythril, echidna, foundry, solhint · openssl, john, hashcat, radare2, apktool, jadx, exiftool, binwalk, yara (48) - approval-gated post-ex drivers (opt-in + human approval + local CLI installed) — metasploit (
msfconsole, riskTierdangerous), hydra (credential); bloodhound is import-only (graph pipeline) - burp suite integration (genuinely not integrated)
7. Pliny Specials — ⛔ RETIRED
⛔ These nine tools were REMOVED and are NOT shipped or callable. The
/api/pliny/*routes and thepliny_*MCP tools no longer exist in the codebase (see the retirement notes later in this file). Onlysecurity_reconsurvives as a live MCP tool. The table and feature checklists below are kept only as a historical record of the retired design — every[x]describes what the removed design once did, not a shipping capability. Nothing in this section is wired up.
The Nine Pliny Specials (all retired)
| Tool | Power | Type | Status |
|---|---|---|---|
| LEVIATHAN | 99 | Kill Chain Orchestrator | ⛔ RETIRED |
| SPHINX | 88 | Vulnerability Validator | ⛔ RETIRED |
| GORGON | 92 | Precision Exploitation | ⛔ RETIRED |
| CERBERUS | 85 | Privilege Escalation | ⛔ RETIRED |
| TYPHON | 90 | Payload Encoding | ⛔ RETIRED |
| GRIFFIN | 95 | Secret Harvesting | ⛔ RETIRED |
| SIMURGH | 100 | Zero-Day Research | ⛔ RETIRED |
| HYDRA | 85 | Multi-Vector Attacks | ⛔ RETIRED |
| ARACHNE | 87 | Exploit Chaining | ⛔ RETIRED |
The per-tool checklists below document the retired design (historical only):
LEVIATHAN Features
- Engagement planning
- Phase orchestration
- Tool recommendations
- Autonomy levels (FULL, GUIDED, MANUAL)
- Real-time execution
- Progress tracking
- Abort/rollback capability
SPHINX Features
- Finding analysis
- False positive elimination
- PoC generation
- Confidence scoring
- Integration with major scanners
- Automated retesting
- Finding deduplication
GORGON Features
- Payload selection
- Delivery method planning
- Stealth rating calculation
- Mode selection (surgical, aggressive, stealth)
- Actual exploit execution
- Session management
- Post-exploitation hooks
TYPHON Features
- URL encoding (single/double)
- Base64 encoding
- Hex encoding
- Unicode escaping
- HTML entity encoding
- WAF bypass techniques
- Custom encoding chains
- Context-aware encoding
- Encoding detection/decode
GRIFFIN Features
- JWT extraction and decoding
- AWS credential detection
- GitHub token detection
- API key extraction
- Password pattern matching
- Private key detection
- Cloud credential validation
- Token expiry checking
- Credential spraying integration
SIMURGH Features
- Vulnerability class targeting
- Protection analysis
- Attack surface mapping
- Exploitation primitive identification
- Fuzzer integration
- Crash analysis
- Exploit development assistance
CERBERUS Features
- Linux privesc vectors (sudo, SUID, kernel, cron)
- Windows privesc vectors (UAC, services, tokens)
- Confidence scoring
- Recommended execution order
- Automated enumeration
- Exploit suggestion
- Success probability calculation
HYDRA Features
- Multi-vector coordination with 14 attack vectors
- Parallel attack planning with configurable heads
- Vector selection (SQLi, XSS, SSRF, SSTI, LFI, RCE, etc.)
- Priority scoring based on target type
- API integration for enhanced analysis
- Web UI integration with real-time status
- Comprehensive payload databases per vector
- Rate limiting
- Result aggregation
- Attack surface coverage metrics
ARACHNE Features
- Chain building from vulnerabilities
- Objective-based pathfinding (14 attack primitives)
- 8 strategic objectives with MITRE ATT&CK mapping
- 18 known exploit chains with CVE mappings
- Reliability and detection risk scoring
- Playbook generation with phases
- API integration for enhanced chains
- Web UI
findChainsEnhanced()method - Chain length limits
- Multiple chain generation
- Graph-based visualization
- Chain validation
8. Evidence Vault
Finding Management
- Create findings with severity
- CVSS score support
- Verification status tracking
- Query by severity/target/operator
- Evidence attachment
- Finding templates
- Duplicate detection
- Finding workflows (triage → verify → report)
Credential Management
- Credential storage
- Type classification (password, hash, token, key, certificate)
- Source tracking
- Query all credentials
- Credential validation
- Password cracking integration
- Credential rotation detection
Evidence Types
- Screenshots
- HTTP requests/responses
- Command output
- Files
- Network captures
- Video recordings
- Memory dumps
- Disk images
Severity Levels
- Critical (score: 10)
- High (score: 7.5)
- Medium (score: 5)
- Low (score: 2.5)
- Info (score: 0)
- CVSS to severity conversion
9. OPSEC Layer
OPSEC Levels
- Silent - Maximum stealth, 1 detection limit, 5min cooldown
- Covert - Balanced, 3 detection limit, 1min cooldown
- Loud - Speed priority, 20 detection limit, 2s cooldown
Detection Management
- Detection event recording
- Abort recommendation threshold
- Cooldown enforcement
- IOC (Indicator of Compromise) tracking
- Detection pattern analysis
- Evasion recommendations
- Real-time alert integration
OPSEC Features
- Traffic blending option
- Timing jitter
- Logging sanitization
- Cleanup on complete
- [~] Avoid detection heuristics
- TOR/proxy integration
- Domain fronting
- Traffic encryption
- Decoy traffic generation
OPSEC Presets
-
createSilentOpsecConfig() -
createAggressiveOpsecConfig() -
createBalancedOpsecConfig() -
createAPTOpsecConfig() -
createRedTeamOpsecConfig()
10. Communications
Channel Types
- Broadcast - All operators
- Direct - One-to-one
- Team - Group channels
- Encrypted channels
- External C2 integration
Message Types
- Intel - Intelligence sharing
- Task - Task assignments
- Alert - Warnings
- Status - Status updates
- Finding - Finding reports
- Coordination - Team coordination
- Heartbeat - Health checks
- Emergency - Critical alerts
Comms Features
- Channel creation/closure
- Subscription management
- Message routing
- Message history
- Priority levels
- Message encryption
- Message acknowledgment
- Offline message queue
11. Analysis & Reporting
Report Types
- Executive summary
- Technical findings
- Full report
- Findings only
- Compliance report (PCI, HIPAA, SOC2)
- Remediation report
- Trend analysis
Report Features
- Risk rating calculation
- Attack path inference
- Recommendation generation
- Markdown export
- [~] HTML export
- PDF export
- DOCX export
- JSON/XML export
Analysis Features
- Severity aggregation
- Target coverage analysis
- Finding categorization
- CVSS calculation
- Risk scoring models
- Trend analysis over time
- Comparison with baselines
Executive Summary Includes
- Total findings by severity
- Risk rating (Critical/High/Medium/Low)
- Top recommendations
- Key findings highlights
- Business impact assessment
- Compliance status
- Remediation timeline
12. CLI Interface
Commands
-
t3mp3st- Interactive mode (default) -
t3mp3st setup- Configuration wizard -
t3mp3st status- Show configuration -
t3mp3st test- Test LLM connection -
t3mp3st models- List available models -
t3mp3st run <mission>- Run mission file -
t3mp3st report <output>- Generate report -
t3mp3st import <file>- Import targets -
t3mp3st export <file>- Export findings
Interactive Mode Features
- Start new operation (4 presets)
- View status (operators, findings, OPSEC)
- Spawn operator (8 archetypes)
- Add target
- Create mission
- Chat with AI
- Generate report
- Stop operation
- Settings management
- Mission templates
- Batch operations
- Scripting support
UI Features
- Colored output (chalk)
- ASCII banner (figlet)
- Gradient text
- Progress spinners (ora)
- Interactive prompts (inquirer)
- Table display (cli-table3)
- Box formatting (boxen)
13. API Server
Server Configuration
- Express.js server
- Port 3333 (configurable)
- CORS enabled
- JSON body parsing
- Static file serving for UI
Health Endpoints
-
GET /api/health- Server health -
GET /api/llm/status- LLM connection status
Pliny Specials Endpoints — ❌ RETIRED (removed 2026-06)
The /api/pliny/* "adversarial engine" routes were removed. They were static payload-catalog
lookups with no measured uplift to the benchmark and were never on the hunt path. The live offensive
surface is the mission-control API (/api/mission/*), the arsenal (/api/tools/*), and the recon
engine — documented in the sections above. No /api/pliny/* route exists in the server today.
Tool Endpoints
-
POST /api/tools/execute- Execute whitelisted tool -
POST /api/tools/recon- Quick reconnaissance -
GET /api/tools- List available tools -
GET /api/tools/:id/history- Tool execution history
LLM Endpoints
-
POST /api/llm/chat- Chat with LLM -
POST /api/llm/stream- Streaming chat -
GET /api/llm/models- List models
Missing Endpoints (Planned)
-
GET /api/operators- List operators -
POST /api/operators- Spawn operator -
DELETE /api/operators/:id- Remove operator -
GET /api/targets- List targets -
POST /api/targets- Add target -
GET /api/findings- List findings -
GET /api/missions- List missions -
POST /api/missions- Create mission -
GET /api/reports- Generate report
14. MCP Server (Agent Integration)
MCP Tools Exposed
The pliny_* MCP tools were retired (2026-06) alongside the /api/pliny routes — they were
catalog lookups, not engines. The MCP server exposes the real, tool-backed surface:
-
security_recon- Network reconnaissance (nmap / DNS / HTTP / fingerprinting, every finding traces to live tool output)
MCP Features
- Tool definitions with JSON Schema
- Tool execution handlers
- Stdio transport
- Error handling
- SSE transport
- WebSocket transport
- Authentication
- Rate limiting
- Usage tracking
Integration Support
- Claude Desktop configuration
- Claude Code integration
- ChatGPT plugin format
- LangChain tools
- AutoGPT plugins
15. Web UI Dashboard
Navigation
- Sidebar navigation
- Section collapse/expand
- Mobile responsive toggle
- Active section highlighting
- Badge indicators
Dashboard Page
- Kill chain visualization
- Operator swarm status
- Live intel feed
- Statistics cards
- OPSEC monitor
- Quick actions
Pages Implemented
- Dashboard - Overview
- Operators - Agent management
- Missions - Mission control
- Evidence - Findings vault
- Arsenal - Tool registry
- Terminal - Command execution
- Benchmarks - Performance testing
- Config Library - OPSEC presets
- CTF Range - Practice targets
- Settings - Configuration
- About - Project info
Visual Features
- Dark cyberpunk theme
- Neon color scheme
- Animated elements
- Grid background
- Progress bars
- Status indicators
- Code syntax highlighting
Interactive Features
- Real-time updates
- Filter controls
- Modal dialogs
- Toast notifications
- Keyboard shortcuts
- Copy to clipboard
- View source on GitHub links
- Drag and drop
- Export functionality
- Dark/light theme toggle
Arsenal Features
- Tool cards with stats
- Category filtering
- Search functionality
- Tool info modals
- Code viewer tab
- Stats tab
- Run demo buttons
- Copy code functionality
- View full source links
16. Advanced Modules
Module Implementations
| Module | Purpose | Status |
|---|---|---|
| KnowledgeBase | CVE/MITRE database | [x] Full - 20 CVEs, 40+ MITRE techniques, 15 patterns |
| EvasionEngine | Encoding & bypass | [x] Full - 10 encoders, obfuscators, WAF bypass |
| ExploitEngine | Payload generation & delivery | [~] Stub |
| ScannerOrchestrator | Multi-scanner coordination | [~] Stub |
| BrowserAutomation | Headless browser control | [~] Stub |
| BenchmarkRunner | Performance testing | [~] Stub |
| ReasoningEngine | Chain-of-thought reasoning | [~] Stub |
| CognitionEngine | Advanced thought patterns | [~] Stub |
| SwarmController | Multi-agent swarm behavior | [~] Stub |
| CloudSecurityEngine | Cloud enumeration | [~] Stub |
| PersistenceController | Implant deployment | [~] Stub |
| LearningEngine | Experience recording | [~] Stub |
| ProtocolHandler | Protocol abstraction | [~] Stub |
| ReportingEngine | Automated reports | [~] Stub |
| WorkflowOrchestrator | Multi-step automation | [~] Stub |
KnowledgeBase Features (NEW)
- 20 critical CVE entries (Log4Shell, Spring4Shell, ProxyLogon, etc.)
- 40+ MITRE ATT&CK techniques across all tactics
- 15 vulnerability patterns for injection detection
-
query()- Search CVEs, techniques, patterns -
getCVE()/getTechnique()- Direct lookups -
getTechniquesByTactic()- Filter by MITRE tactic -
getCriticalCVEs()- Filter by CVSS threshold -
matchPatterns()- Detect vulnerabilities in content
EvasionEngine Features (NEW)
- 10 encoding schemes: base64, base64url, hex, unicode, rot13, url, html, octal
- Multi-layer encoding with
multiEncode() - Code obfuscators: string_split, variable_rename, junk_code
- Sandbox detection (VM, timing, CPU, memory checks)
-
generatePolymorphic()- Multiple payload variants -
getWAFBypassVariants()- WAF evasion techniques
Planned Advanced Features
Cognition Patterns
- Chain-of-Thought (CoT) reasoning
- ReAct (Reasoning + Acting)
- Tree-of-Thought exploration
- Self-consistency checking
- Debate mode (multiple perspectives)
Swarm Intelligence
- Pheromone-based communication
- Emergent behavior coordination
- Task distribution algorithms
- Collective decision making
Cloud Security
- AWS enumeration & exploitation
- GCP security testing
- Azure penetration testing
- Multi-cloud support
- IAM analysis
- S3/blob storage auditing
Persistence & C2
- Implant generation
- C2 channel establishment
- Beacon management
- Persistence mechanism deployment
Learning & Adaptation
- Experience recording
- Pattern recognition
- Technique effectiveness tracking
- Adaptive strategy selection
17. Configuration System
Storage
- Persistent config via Conf library
- Project-scoped storage
- Environment variable loading
- .env file support
- Config file import/export
- Config encryption
- Remote config sync
API Key Management
- Store/retrieve API keys
- Key validation
- Multi-provider support
- Environment variable precedence
- Key rotation
- Key expiry warnings
- Secure key storage (keychain)
Provider Configuration
- Base URL customization
- Default model selection
- Site URL/name for OpenRouter
- Timeout configuration
- Temperature settings
- Max tokens configuration
OPSEC Configuration
- Level presets
- Detection thresholds
- Cooldown settings
- Cleanup preferences
UI Preferences
- Show/hide banner
- Color output toggle
- Verbose logging toggle
- Theme selection
- Custom shortcuts
18. Integration & Extensibility
Current Integrations
- OpenRouter API
- Anthropic API
- OpenAI API
- Ollama (local)
- Model Context Protocol (MCP)
- LangChain
- AutoGPT
- BabyAGI
Extensibility Points
- Custom tools via Arsenal.register()
- Custom operators via createOperator()
- Event hooks via EventEmitter
- Custom LLM providers via adapters
- Plugin system
- Webhook notifications
- External C2 integration
- SIEM integration
Export Formats
- JSON (findings, reports)
- Markdown (reports)
- HTML
- CSV
- SARIF (Static Analysis Results)
- STIX/TAXII (threat intelligence)
Import Formats
- Nmap XML
- Nessus reports
- Burp XML
- OWASP ZAP
- Nuclei JSON
- Shodan export
Feature Roadmap Priority
P0 - Critical (Next Release)
- Full API endpoint coverage for operators/targets/missions
- Real exploit execution in Pliny tools
- PDF report generation
- Scanner import (Nmap, Nuclei)
P1 - High (Near Term)
- Cloud security module (AWS/GCP/Azure)
- Function calling in LLM backbone
- WebSocket real-time updates
- Plugin system architecture
P2 - Medium (Mid Term)
- Cognition patterns (CoT, ReAct)
- Swarm intelligence
- Learning engine
- Distributed operation support
P3 - Low (Long Term)
- Mobile app
- SaaS deployment
- Marketplace for tools/plugins
- AI-powered remediation suggestions
Metrics
Total Features Tracked: ~300
- Implemented: ~150 (50%)
- Partially Implemented: ~30 (10%)
- Planned: ~120 (40%)
Last updated: 2026-07 Version: 1.0.0