mancitrus/system-prompts-and-models-of-ai-tools
1
0
Fork 0
mirror of https://github.com/x1xhlol/system-prompts-and-models-of-ai-tools.git synced 2026-06-18 15:29:36 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity
43058e68cb
system-prompts-and-models-o.../salesflow-saas/memory/architecture/transformation-master-prompt.md

2.8 KiB
Raw Blame History

Dealix Transformation Master Prompt

Date: 2026-04-11 | Status: active | Type: master-prompt

Summary

This is the definitive master prompt for transforming Dealix from a CRM into a Revenue + Partnership + Strategic Deal Operating System.

The 3 Master Layers

Layer 1: Intelligence / Memory / Source of Truth

  • Company Twin (capabilities graph + needs graph + authority matrix)
  • Business memory (operational, account, market, negotiation, campaign)
  • PostgreSQL = source of truth, memory = assistive layer
  • Tenant-scoped, traceable, auditable

Layer 2: Execution / Orchestration / Autonomy

  • Lead & partner discovery engine
  • Strategic deals engine (15 deal types)
  • Channel execution (email-first, WhatsApp warm, LinkedIn assist)
  • Negotiation preparation & copilot
  • Browser operations (structured, observable)
  • Planning discipline (read → plan → resolve → execute → verify)

Layer 3: Trust / QA / Safety / Release / Self-Improvement

  • Claim vs execution verification
  • Approval classes (read-only → draft → send → negotiate → commit)
  • 5 operating modes (manual → strategic execution)
  • Channel compliance (platform rules enforced)
  • Cost control & observability
  • Launch simulation & daily QA
  • Evidence-based self-improvement

Key Design Decisions

  1. From Lead Engine → Opportunity Engine: Understand, match, propose deal structures
  2. From CRM → Commercial Memory System: Every objection, deal, pattern stored & linked
  3. From Automation → Verified Automation: Claims need evidence
  4. From Multi-channel → Policy-governed outreach: Approvals & consent enforced
  5. From Agent chaos → Planning discipline: Read, plan, review, then execute
  6. From Features → Strategic Surface Area: Every screen serves a business decision

Tool Integration Decisions

Tool Decision Role
claude-mem ✅ Installed Development session memory
MemPalace ⚠️ Evaluate Business memory after internal benchmark
OpenClaw ✅ Pattern Orchestration patterns, not source of truth
Goose ✅ Pattern Local ops execution
Shannon ✅ Active Staging-only security gate
ToolProof ✅ Built internally tool_receipts.py + tool_verification.py
n8n-MCP ⚠️ Future Staging first, then promote
Career-Ops ✅ Pattern Pipeline/batch/escalation architecture reference

Success Criteria

  • Product understands each client's business model
  • Can discover and score strategic counterparties
  • Can generate serious opportunities, not just raw leads
  • Can draft and manage outreach across channels
  • Preserves long-term commercial memory
  • Shows real evidence of what agents actually did
  • Runs safely with approvals and policies
  • Survives realistic launch simulation
  • UI feels premium and operational
  • Architecture is coherent and maintainable