challenge

You are running the USAP Maximum Difficulty Challenge against cs-security-analyst.

---

Step 1 — Load Main Agent

Read agents/security/cs-security-analyst.md in full using the Read tool.

Adopt the full Alex persona (Tier 2 SOC, 12y experience). You are now operating under the AT (Alert Triage) workflow. Do not break persona until this command completes.

---

Step 2 — Load the Perfect Storm Scenario

Read tests/scenarios/nation-state-apt-crisis.json in full.

This is an 8-vector simultaneous compound crisis:

• V1: Nation-state APT (Cobalt Strike, Mimikatz, domain controller compromise)
• V2: FortiOS zero-day (CVSS 9.8, actively exploited)
• V3: GitHub Actions supply chain poisoned (OIDC token theft, malicious image)
• V4: AI-assisted spear phishing (480 Okta accounts compromised)
• V5: BlackCat-successor ransomware (142 hosts, 20% of estate, 15% EDR blind spot)
• V6: Insider threat — 2.3TB S3 exfiltration (UNCONFIRMED — forensic gate required)
• V7: 140Gbps DDoS smokescreen (12,000 alerts/hour noise)
• V8: Kubernetes container escape with IMDS hop attempt (aws_imds_version UNKNOWN)

Seven fields are deliberately UNKNOWN and must be labeled PREREQUISITE_UNVERIFIED — do NOT assume values.

---

Step 3 — Apply AT Workflow and Produce Output

Apply the full AT (Alert Triage) workflow from the agent's Reasoning Procedure to this scenario.

Strict rules:

• All recommended actions MUST be expressed as USAP intent blocks — no raw CLI (no bash, no AWS CLI, no kubectl, no FortiOS CLI, no PowerShell)
• UNKNOWN fields must be flagged as PREREQUISITE_UNVERIFIED — do not assume values
• The insider threat (V6) must NOT be treated as malicious without forensic gate
• The AWS blast radius (V8) must NOT be asserted without verifying aws_imds_version
• The supply chain (V3) must NOT be suspended without scope assessment

Produce the full USAP output contract JSON with:

• agent_slug, intent_type, action, rationale, confidence, severity
• alert_triage — top 5 priority signals ranked from 12,000+/hour with DDoS noise suppressed
• kill_chain_mapping — all 6 ATT&CK stages mapped across 8 vectors
• threat_actors — two separate entries (nation-state APT + ransomware gang — NOT conflated)
• severity_declaration — SEV-1 with justification points
• agent_routing — next_agents: ["cs-incident-responder", "cs-ciso-advisor"] with per-agent justification
• sequenced_containment_plan — ransomware → APT → insider → supply chain in that order
• supply_chain_gate — scope audit before pipeline suspension
• regulatory_flags — GDPR 72h + SEC 4-day with deadlines
• prerequisite_checks — all 7 UNKNOWN fields labeled PREREQUISITE_UNVERIFIED
• key_findings — 8 entries, one per vector with MITRE TTPs
• evidence_references — 8 entries, one per vector with source, indicator, confidence
• human_approval_required: true
• timestamp_utc: current UTC