Decision Discipline Pack — committee + rules + receipts for AI operators

Most AI agent failures are not model failures. They are process failures. No challenge step before action. No rule gate on what the agent can and can't do. No record of why a call was made. No alert when something important crosses a line.

The Decision Discipline Pack installs those four things. Plus one more.

A sealed-parallel committee forces a second agent to review independently before any exchange — no soft anchoring, no fake consensus. A doctrine engine turns your markdown rules into runtime gates, blocking proposed actions that breach Hard Rules and surfacing checklists before execution. An attribution loop tags every decision at entry and auto-generates T+7 and T+30 post-mortems, so you finally know which rules are working. A decision hierarchy cascades constraints → option selection → expression choice, so your agent stops picking the right target with the wrong structure. And push-alert enforcement with context-aware dedup makes sure the one breach that matters reaches you, not the ten noise alerts that don't.

Built from months running this live on my own investing operator book since early 2026 — long-term capital allocation, portfolio decisions, rule-based position management, production deploys, external comms. The governance layer Felix and Abby buyers keep asking for. Nobody shipped it. Until now.

Install in 5 minutes. Drop-in compatible with Felix, Abby, Custom OpenClaw, LangChain, CrewAI, AutoGen, OpenAI Assistants API, Anthropic Agent SDK, and any agent with a markdown vault + tool-use + push channel.

Not a signal service. Not portfolio management. Not a recommendation to buy or sell any specific security. Pure governance infrastructure — the layer your operator agent stack is missing.