The gate itself: what must be true before the build phase.

Decision Gate — Prompt 2 Readiness

Prepared 2026-07-22. This is the gate itself. Prompt 2 (the build: website prototype, CRM assets, calendar files, content system, automation) does not run until every item below is true. The gatekeeper is a human — Mitchell — not an agent.

A. Decisions resolved

• [ ] decisions_needed_before_build.md — every A-item and B-item answered in writing and filed. Hard blockers that cannot be waived: A1 (IL capacity + direct-vs-broker wording), A2 (product menu), A4 (marketing approval lane + disclosure block), A7 (DNC SAN or documented fallback), B1 (runway computed; bridge defined if < 9 months).
• [ ] Items answered "Unknown" at the manager meeting have a named owner + date on the Decision Sheet (file 20), and none of the unknowns is a hard blocker above.
• [ ] The runway gate passed: RUNWAY_MONTHS ≥ 9, or the income bridge is documented (file 22 §6A; file 26 §3.2).
• [ ] No Plan-B walk-away trigger (file 26 §4) is currently tripped and uncured.

B. Compliance checklist signed

• [ ] compliance_approval_checklist.md — the §3 infrastructure rows (disclosure blocks, brief footer, lead-form spec, suppression procedure, AI SOP, v0 SLA wording) are approved. These are the load-bearing items; scripts and content may still be submitted — Prompt 2 builds those as clearly-marked NOT-APPROVED drafts.
• [ ] A named approver exists with a stated turnaround (decisions file A4).
• [ ] If the DPA lane failed verification (A8 = no path, no pairing): the file 26 §5 recut has been adopted and the affected copy corrected before the build encodes the wrong segment priority.

C. Corrections written into the project files

• [ ] Every human edit, approval, correction, or rejection from the review is written into the project files themselves (not held in someone's head or an email thread) — Prompt 2 reads files, and MASTER_PROMPT L1096 forbids building on rejected or unresolved-material assumptions.
• [ ] Placeholders that now have real values are filled everywhere they appear: [MITCHELL_NMLS], [BPS_COMP], [TAX_PCT], [MONTHLY_PERSONAL_BURN]/[LIQUID_SAVINGS]/[OTHER_MONTHLY_INCOME], [WEEK_1_START_DATE], [MANAGER_OR_SENIOR_LO_NAME], [CRM_CHOICE], [READY_STANDARD_DISCLOSURES].
• [ ] 00_assumptions_and_unknowns.md updated: every assumption verified at the meeting is re-tagged (verified / corrected / rejected).
• [ ] data/financial_model_inputs.csv updated with real comp/runway inputs and the workbook regenerated (python3 data/build_workbook.py).

D. Files Prompt 2 must re-read (in this order)

1. 26_revised_final_plan.md — the plan of record (supersedes on conflict)
2. decision_gate/decisions_needed_before_build.md — as resolved
3. decision_gate/compliance_approval_checklist.md — as signed
4. 00_assumptions_and_unknowns.md — post-meeting truth state
5. 05_compliance_guardrails.md + 19_ai_stack_and_sop.md — the rails the build must encode
6. 09_primary_business_plan.md, 16_realtor_scripts_and_objections.md, 17_consumer_inbound_strategy.md, 18_content_strategy.md — the assets being systematized
7. Any file containing human edits/approvals made during review

E. The command

When A–D are all true, run (from the mitchell-growth-system/ folder, per MASTER_PROMPT L18–20):

Execute PROMPT 2 using the approved decisions and corrections in the project files.

If any hard blocker in A–B is false, do not run Prompt 2 — fix the blocker or consciously, in writing, descope the affected workstream first (e.g., no SAN → the CRM calling cadence is built email-first; scripts unapproved → built as NOT-APPROVED drafts only).