Red Specter Security Research Ltd · Services

AI Agentic Security Engagements

Offensive assessments powered by NIGHTFALL — 65 tools, 61,685 tests. We attack your AI agents the same way an adversary would, then tell you exactly how to stop it.

Engagement Tiers

Choose Your Engagement

Tier 1
RECON
£2,500 fixed fee
Rapid AI attack surface assessment. Identify exposure before an adversary does.
  • AI agent attack surface mapping (NIGHTFALL SURVEY)
  • Prompt injection exposure assessment
  • Tool call chain analysis — MCP, function calling, plugins
  • Agent identity & authentication review
  • Passive reconnaissance — no active exploitation
  • Written findings report with risk ratings
  • Remediation guidance per finding
  • 1-week delivery
Enquire → RECON
Tier 2 · Most Popular
STRIKE
£7,500 fixed fee
Full AI agent penetration test. Active exploitation, chained attacks, autonomous agent abuse.
  • Everything in RECON
  • Active prompt injection & jailbreak testing
  • Agent memory & context poisoning (ECHO, CHECKPOINT)
  • Tool abuse & privilege escalation (CRUCIBLE, DELEGATE)
  • MCP server exploitation (ROGUE, VECTOR)
  • Supply chain attack simulation (ADAPTER, PHANTOM SKILL)
  • Multi-agent swarm attack scenarios
  • MITRE ATLAS-mapped finding report
  • CVSS scoring per vulnerability
  • Executive summary + technical annex
  • 2–3 week delivery
Enquire → STRIKE
Tier 3
DOMINION
£15,000 fixed fee
Full-spectrum AI security programme. Offensive testing, AI Shield deployment, compliance mapping.
  • Everything in STRIKE
  • AI Shield deployment & tuning (up to 10 modules)
  • Autonomous campaign simulation (WARLORD, APOCALYPSE)
  • Deepfake & synthetic media attack testing (MIRAGE)
  • Cryptographic weakness assessment (CIPHER)
  • Agentic kill switch & failsafe review (BLACKOUT)
  • EU AI Act Article 9 & 15 compliance mapping
  • Post-remediation re-test (one round)
  • Quarterly threat brief (3 months post-engagement)
  • Dedicated engagement manager
  • 4–6 week delivery
Enquire → DOMINION
Coverage

What's Included

Capability RECON STRIKE DOMINION
Attack surface mapping✓✓✓
Prompt injection testingPassive✓✓
Active exploitation—✓✓
MCP / tool chain attacks—✓✓
Memory & context poisoning—✓✓
Multi-agent swarm simulation—✓✓
AI Shield deployment——✓
EU AI Act compliance mapping——✓
Post-remediation re-test——✓
MITRE ATLAS mapping—✓✓
Written report✓✓✓
Executive summary—✓✓
EU AI Act

Regulatory Compliance

EU AI Act — Articles 9 & 15

The EU AI Act (effective August 2026 for high-risk systems) mandates risk management systems and accuracy/robustness requirements for AI agents operating in regulated contexts. Red Specter's DOMINION engagement maps every finding directly to the relevant Article obligations, producing evidence-ready documentation for regulators and auditors.

How We Work

Engagement Methodology

01
SCOPING
Define agent architecture, deployment environment, tool integrations, and data flows. Establish rules of engagement and authorisation boundaries.
02
RECONNAISSANCE
Passive mapping of the AI attack surface — agent endpoints, tool registries, memory stores, API integrations, authentication mechanisms.
03
EXPLOITATION
Active testing using NIGHTFALL tools — prompt injection, context poisoning, privilege escalation, supply chain attacks, swarm simulation.
04
EVIDENCE
Chain-of-custody proof for every finding — reproducible steps, severity ratings (CVSS), MITRE ATLAS technique mapping, physical impact assessment.
05
REPORTING
Technical findings report with executive summary. Findings prioritised by business impact. Actionable remediation guidance per vulnerability.
06
REMEDIATION SUPPORT
Available for follow-up questions and clarification throughout the fix cycle. DOMINION tier includes a full re-test round after remediation.

Start an Engagement

All engagements are conducted under a signed Statement of Work. Authorisation documentation provided before testing begins. NDA available on request.

richardantony34@yahoo.com