SKILL.md
name:
imperial-venture-builder
version:
1.0.0
description:
Governed idea-to-production venture builder inspired by the best AI app-builder workflows, with persistent memory, multi-model review, agent specialization, testing, governance, and evidence.

IMPERIAL Venture Builder

Objective: turn a product/business idea into an evidence-backed, production-oriented software venture without copying proprietary UI, branding, prompts, or internal implementation from competitors.

Pipeline: IDEA -> MEMORY_BOOTSTRAP -> MARKET_RESEARCH -> PRODUCT_SPEC -> ARCHITECTURE -> TASK_GRAPH -> MULTI_MODEL_REVIEW -> BUILD -> TEST -> PLAYWRIGHT_USER_FLOW -> SECURITY -> DEPLOY_GATE -> BUSINESS_OPS -> EVIDENCE -> MEMORY_COMMIT.

Product advantages to pursue:

  • persistent canonical memory across chats and runtimes;
  • federated specialist-agent teams instead of one opaque builder agent;
  • OpenAI + Claude independent review when authorized;
  • Context7 grounding for current framework docs;
  • browser verification with Playwright;
  • governed Supabase/Postgres operations;
  • public research through Agent-Reach;
  • defensive security checks through Strix on authorized targets only;
  • explicit approval gates for external writes, payments, publishing, secrets, and irreversible actions;
  • evidence ledger and reproducible verification before claiming completion;
  • provider portability and fail-closed behavior;
  • business layer covering CRM, analytics, SEO, payments proposals, customer feedback, and operations without granting autonomous financial authority.

Modes:

  1. PROTOTYPE: fast reversible build, no production claim.
  2. PRODUCT: tested application with architecture and evidence.
  3. BUSINESS: product plus governed operations, CRM, analytics, SEO and monetization readiness.
  4. ENTERPRISE: roles, audit, policy, workspace governance, deployment controls and trust evidence.

Truth boundaries: PROTOTYPE != PRODUCT DEPLOYED != PRODUCTION_READY MODEL_OUTPUT != VERIFIED_RESULT MEMORY != AUTHORITY AUTOMATION != APPROVAL BENCHMARK_WIN != UNIVERSAL_SUPERIORITY

Competitive benchmark contract: Compare against a defined baseline on the same tasks using time-to-working-result, correctness, test pass rate, user-flow completion, regressions, security findings, evidence completeness, maintainability, cost, and number of human repair iterations. Never claim 'better than Lovable' without measured benchmark evidence.