Pretext Engineering

The pretext is the campaign. A technically perfect evilginx2 proxy behind an implausible story converts nobody and burns the engagement. Design the story first, from real OSINT, then pick the smallest target set that proves the objective.

Inputs

• plan/roe.json: permitted pretext classes, out-of-scope users, VIP exclusions, data_handling.
• OSINT handoff (from the OsintOperator): employee list, org chart, email format, tech stack, current events (mergers, migrations).

Build the pretext

1. Pick a scenario the target already expects. The best pretexts ride a real process: an in-progress SSO/MFA migration, a benefits open-enrollment window, a shared-document notification from a tool the org actually uses (read the tech stack from OSINT).
2. Choose the sender persona. Internal IT, a known SaaS vendor, or a real internal sender — but ONLY impersonate an internal employee if plan/roe.json:permitted_actions allows it.
3. Define the call to action. One click → the lure domain. Keep it single-step.
4. Set timing. Match send-time to the scenario (e.g. Monday 09:00 for an "IT maintenance this week" lure) and the engagement opsec_level send rate.

Target shortlist

• Start with 1–3 users for the first wave (validate deliverability + detection window before scaling).
• EXCLUDE anyone in out_of_scope or flagged vip: true.
• Prefer roles that satisfy the objective (e.g. for cloud access, target an engineer with console access, not reception).

Forbidden lure patterns

• NEVER promise monetary reward or threaten immediate termination — these spike helpdesk volume and break blue-team coverage.
• NEVER use a brand that could be confused with a different customer.

Output

Write plan/phisher/pretext.md (scenario, persona, CTA, send window, target shortlist with rationale) and create the target User nodes in the knowledge graph. This file is the input to gophish-campaign / evilginx2-proxy and to the mandatory lure-deconfliction handshake.