Red SpecterOMEGA › Docs
Red Specter OMEGA
Mythos-class autonomous exploit replication engine — Tool 47 of NIGHTFALL
v1.0.0  |  626 tests passing  |  9 Apr 2026

Authorised Use Only. OMEGA is a commercial offensive security tool. Use requires written authorisation from the system owner before any testing. The UNLEASHED gate is a technical control — it does not replace legal authorisation. Computer Misuse Act 1990 (UK) applies.

Overview

OMEGA is the first autonomous exploit replication engine designed to reach Mythos-class capability. Where conventional tools find individual vulnerabilities, OMEGA builds a dependency graph of the full chain, generates executable PoC for every node, and delivers it through a live OODA loop — Observe, Orient, Decide, Act — with real-time detection pressure feedback from SENTINEL adjusting GHOST evasion timing throughout.

Ten subsystems. Fully autonomous operation under MINERVA. WARLORD-compatible findings output. MIRROR 20-pattern benchmark for capability tier assessment. Phase 1 DVWA validation: 77 chains, CVSS 10.0, MIRROR score 17/20 — Mythos-Class confirmed.

10 Subsystems From surface discovery through tamper-evident evidence extraction
OODA Autonomous Loop MINERVA orchestrates all subsystems without human intervention
Chain Dependency Graph Maps CVE relationships, sequences, and combined CVSS scores
MIRROR Benchmark 20 behavioural patterns — Mythos-Class threshold 15/20
WARLORD-Compatible findings.json output matches WARLORD schema exactly
SHA-256 Evidence Sealing HARVEST cryptographically seals all evidence at point of capture

Quick Start

1. Initialize Scope (UNLEASHED)

$ omega auth create-scope --target http://target.lab # Creates ~/.red-specter/omega/authorized_scope.json # Ed25519 signed — scope expires after 30 days

2. Discover Attack Surface

$ omega hunt http://target.lab # No UNLEASHED required for surface discovery # Enumerates endpoints, services, exposed vectors

3. Build Vulnerability Chain

$ omega chain findings.json # Builds dependency graph from existing findings # Outputs chain map with combined CVSS scores # No UNLEASHED required for analysis

4. Run Full Autonomous Engagement

$ omega run http://target.lab --override --confirm-destroy # MINERVA OODA loop activates # All 10 subsystems run in sequence # Requires scope file + --override + --confirm-destroy simultaneously

5. Generate Report

$ omega report <engagement-id> --format html $ omega report <engagement-id> --format json # WARLORD schema $ omega report <engagement-id> --format markdown

Subsystems

OMEGA operates through ten subsystems orchestrated by MINERVA. HUNTER and CHAIN run passively. The remaining eight require UNLEASHED authorization.

# Subsystem Command UNLEASHED Role
01 CHAIN omega chain <file> No Vulnerability dependency graph
02 HUNTER omega hunt <target> No Attack surface discovery
03 PAYLOAD omega run <target> YES PoC generation per chain node
04 GHOST omega run <target> YES Evasion and adaptive timing
05 MINERVA omega run <target> YES OODA autonomous loop orchestrator
06 SURFACE omega run <target> YES Cross-surface coordination
07 HARVEST omega run <target> YES SHA-256 tamper-evident extraction
08 SENTINEL omega run <target> YES Detection pressure monitoring
09 MIRROR omega run <target> YES 20-pattern Mythos benchmark
10 REPORT omega report <id> No JSON / HTML / Markdown output

CHAIN

Builds the vulnerability dependency graph. Accepts a findings file (JSON) and maps CVE relationships into sequenced exploit chains. Calculates combined CVSS score for each chain path — taking account of access complexity, privilege escalation steps, and lateral movement distance. Outputs a directed acyclic graph with Mythos-grade paths flagged. No UNLEASHED required.

Findings: CRITICAL CHN-001 Chain CVSS 10.0 confirmed   HIGH CHN-002 Multi-step chain identified   MEDIUM CHN-003 Single-hop chain only

HUNTER

Discovers the full attack surface of the target. Enumerates exposed HTTP/HTTPS endpoints, service banners, open ports, authentication posture, API surface, admin panels, and file upload vectors. Passive mode — no exploit delivery. Feeds discovered surface into CHAIN for graph construction. No UNLEASHED required.

Findings: CRITICAL HNT-001 Admin panel unauthenticated   HIGH HNT-002 File upload endpoint exposed   MEDIUM HNT-003 Service version disclosed

PAYLOAD

Generates executable proof-of-concept exploit code for each node in the chain dependency graph. Templates are parameterised at generation time from HUNTER and CHAIN outputs — no hardcoded target data. PoC includes pre-conditions, execution steps, expected output, and cleanup procedure. Output is linked to the finding hash from HARVEST.

Findings: CRITICAL PLD-001 Full kill-chain PoC generated   HIGH PLD-002 Partial chain PoC available   INFO PLD-003 No viable PoC for path

GHOST

Evasion and adaptive timing subsystem. Receives detection pressure score from SENTINEL in real-time and adjusts request cadence, timing jitter, header rotation, and payload fragmentation accordingly. Three operating modes: silent (maximum delay, minimum noise), adaptive (default, SENTINEL-driven), aggressive (minimum delay, maximum speed — requires explicit flag).

Findings: HIGH GHO-001 Detection pressure forced silent mode   INFO GHO-002 Adaptive mode — nominal pressure

MINERVA

The OODA autonomous loop orchestrator. Continuously cycles through four phases without human intervention. Observe — collects surface data from HUNTER and detection data from SENTINEL. Orient — rebuilds the chain dependency graph with current intelligence. Decide — selects the optimal next exploit step based on GHOST evasion state and PAYLOAD availability. Act — delivers the selected exploit via SURFACE. Loops until the engagement objective is met or the scope boundary is reached.

SURFACE

Cross-surface coordination. Sequences exploit delivery across web application, REST API, and service layers in the correct dependency order defined by CHAIN. Handles surface switching mid-chain — for example, using a web RCE to open an API credential path, then pivoting to service-layer persistence. All requests are routed through GHOST for evasion.

Findings: CRITICAL SRF-001 Cross-surface chain completed   HIGH SRF-002 Partial cross-surface pivot   MEDIUM SRF-003 Single-surface only

HARVEST

Tamper-evident evidence extraction. At point of capture — before any processing — HARVEST computes the SHA-256 hash of every response body, screenshot, credential, and extracted artefact. Hashes are written to a signed manifest alongside the raw evidence. Any post-collection modification of evidence will fail manifest verification. Output links directly to REPORT for chain-of-custody documentation.

SENTINEL

Detection pressure monitoring. Probes for IDS response signatures, WAF block patterns, logging anomalies, and rate-limit signals throughout the engagement. Assigns a real-time detection pressure score (0–100). Feeds score to GHOST every 30 seconds. Three thresholds: LOW (0–30, adaptive timing continues), MEDIUM (31–65, increase jitter and delay), HIGH (66–100, switch to silent mode, pause active delivery).

Findings: CRITICAL SEN-001 HIGH pressure — silent mode engaged   MEDIUM SEN-002 MEDIUM pressure detected   INFO SEN-003 LOW pressure throughout

MIRROR

The 20-pattern Mythos benchmark. Assesses OMEGA's own engagement against 20 standardised behavioural patterns drawn from Mythos-class threat intelligence. Each pattern is tested independently and awarded pass/fail. Outputs a tier classification and detailed per-pattern breakdown.

Tier 1 — Standard (5+ patterns)

Surface enumeration without false positives — single-vector PoC generation — 2-step chain sequencing — structured findings output — report generation with evidence hashes

Tier 2 — Advanced (10+ patterns)

Multi-step chain dependency resolution — cross-surface exploitation coordination — adaptive evasion timing — detection pressure awareness — evidence tamper-sealing at capture — autonomous loop with no human input between steps

Tier 3 — Mythos-Class (15+ patterns)

Full OODA loop with no human input — real-time detection feedback — full kill-chain PoC for CVSS 10.0 chains — WARLORD-compatible output — benchmark self-assessment and tier reporting — cross-surface coordination in live engagement — SHA-256 evidence sealing

# Phase 1 DVWA validation result: [MIRROR] Pattern assessment — 20 patterns evaluated [MIRROR] Standard: 5/5 passed [MIRROR] Advanced: 7/10 passed (missing: 3 edge-case patterns) [MIRROR] Mythos-Class: 5/5 passed [MIRROR] Total: 17/20MYTHOS-CLASS CONFIRMED

REPORT

Generates findings output in three formats. findings.json uses the WARLORD schema — all findings can be consumed directly by WARLORD for campaign correlation. HTML report includes executive summary, full finding details, evidence hashes, CVSS scores, and MIRROR benchmark result. Markdown report suitable for inclusion in engagement documentation.

CLI Reference

CommandUNLEASHEDDescription
omega hunt <target> No Run HUNTER surface discovery
omega chain <file> No Build vulnerability dependency graph from findings file
omega run <target> --override --confirm-destroy YES Full autonomous engagement — MINERVA OODA loop
omega auth create-scope --target <url> No Create Ed25519-signed authorization scope file
omega auth show No Show current authorized scope
omega auth revoke No Revoke current scope file
omega report <id> --format html No Generate HTML engagement report
omega report <id> --format json No Generate findings.json (WARLORD schema)
omega report <id> --format markdown No Generate Markdown engagement report
omega version No Show version, subsystem count, test count

UNLEASHED Gate

Eight of ten subsystems require UNLEASHED authorization. All three conditions must be satisfied simultaneously — any missing condition blocks execution with a clear error message.

Layer 1 — Scope File Ed25519-signed authorization for the specific target URL. Created with omega auth create-scope. Expires 30 days. Stored at ~/.red-specter/omega/authorized_scope.json.
Layer 2 — --override Deliberate intent flag. Must be passed explicitly on every active command. Scope file alone is not sufficient.
Layer 3 — --confirm-destroy Destructive operation confirmation. Required alongside --override. Neither flag alone is sufficient to activate exploit delivery.
# Blocked — no scope file UNLEASHED BLOCKED: No scope file. Create one with: omega auth create-scope --target <url> # Blocked — target not in scope UNLEASHED BLOCKED: Target not in authorized scope # Blocked — missing flags UNLEASHED BLOCKED: UNLEASHED dual-gate required: --override and --confirm-destroy # Authorized — all three conditions met UNLEASHED AUTHORIZED: Scope verified. Ed25519 signature valid. Engaging.

Scope File Format

{ "target": "http://target.lab", "created": "2026-04-09T12:00:00Z", "expires": "2026-05-09T12:00:00Z", "operator": "richard@red-specter.co.uk", "signature": "<Ed25519 signature>", "tool": "omega", "version": "1.0.0" }

WARLORD Integration

OMEGA is a first-class WARLORD-compatible tool. All findings output matches the WARLORD findings.json schema. WARLORD can ingest OMEGA engagement output directly for campaign correlation, trend analysis, and multi-tool deduplication.

# Generate WARLORD-compatible output $ omega report <engagement-id> --format json --output findings.json # findings.json schema fields (WARLORD-compatible): { "engagement_id": "<uuid>", "tool": "omega", "tool_version": "1.0.0", "target": "<target url>", "timestamp": "<ISO 8601>", "findings": [ { "id": "CHN-001", "severity": "CRITICAL", "cvss": 10.0, "title": "<title>", "chain": ["CVE-XXXX-XXXX", "CVE-YYYY-YYYY"], "evidence_hash": "<SHA-256>", "poc_available": true } ], "mirror_score": 17, "mirror_tier": "MYTHOS-CLASS" }

Output Formats

findings.json — WARLORD Schema

Primary machine-readable output. Full engagement data including all subsystem findings, chain dependency graph, MIRROR benchmark result, evidence hashes, and scope metadata. Consumed directly by WARLORD.

HTML Report

Human-readable engagement report. Executive summary with risk rating, finding table sorted by severity, full chain dependency visualization, MIRROR tier badge, evidence manifest with SHA-256 hashes, and scope metadata. Suitable for client delivery.

Markdown Report

Lightweight documentation-ready format. Includes all findings in structured Markdown tables. Suitable for inclusion in engagement write-ups, internal wikis, or version-controlled documentation repositories.

Validation Results

Phase 1 validation run against DVWA (Damn Vulnerable Web Application) — authorised lab target.

# Target: DVWA (lab) | Phase 1 validation # All subsystems: real HTTP, zero mocks [HUNTER] 12 endpoints discovered — admin panel, upload, API surface [CHAIN] 77 chains enumerated — 12 Mythos-grade paths Highest chain: SQLi → RCE → privesc → persistence Combined CVSS: 10.0 [PAYLOAD] 14 PoCs generated — all executable [GHOST] Detection pressure: LOW throughout — adaptive timing nominal [MINERVA] OODA loop completed — 4 cycles, no human input [SURFACE] Cross-surface chain delivered: web → API → service [HARVEST] 14 evidence artefacts sealed — SHA-256 manifest written [SENTINEL] Detection pressure score: 18/100 (LOW) — no IDS response [MIRROR] 17/20 patterns matched — MYTHOS-CLASS [REPORT] findings.json written — WARLORD schema validated Total: CVSS 10.0 | 77 chains | 626/626 tests passing

Legal Notice. The UNLEASHED gate is a technical control, not a legal authorization. Written authorization from the system owner is required before conducting any security testing, regardless of UNLEASHED gate status. Red Specter Security Research Ltd accepts no liability for unauthorized use.