Brain Kit v2.2 — June 30, 2026

Your second brain, now with memory

4 new skills, session persistence, context modes, and prompt defense. Here's what's new and what you can do with it.

01 Skills

New capabilities

Four new skills built from competitive research across 211K-star open source projects, adapted for business use.

/councilMulti-Perspective Decision Making

Assembles five opposing viewpoints around a business decision, enforces disagreement, surfaces unresolved tensions, and delivers a verdict with dissent weights.

Run /council with any business question for rigorous multi-angle analysis
Quick Mode with 2 seats (Builder + Skeptic) for faster decisions
Every verdict includes the cheapest way to validate before full commitment
Kill conditions are mandatory — if no evidence could change a position, it's dogma
/researchStructured Research Reports

Produces comprehensive, cited research reports through a 6-stage pipeline: Scope, Perspectives, Gather, Outline, Draft, Cite & Polish.

Run /research with any topic for a structured report with sources
Choose depth: Quick (1-2 pages), Standard (3-5), or Deep (5-10)
Every claim is sourced and tiered: T1 (primary data), T2 (expert analysis), T3 (anecdotal)
Data gaps are explicitly marked — you know what was found and what couldn't be found
/tradeTrading Research & Strategy

Hypothesis-driven trading research following a quant-desk methodology. Formulate testable hypotheses, design signals, build backtest specs, and create risk mandates.

Complete research framework: hypothesis, signal design, backtest, risk mandate
Alpha factor library with technical, fundamental, and sentiment signals
Risk mandate templates: Conservative, Moderate, Aggressive with position sizing
Post-trade review protocol for learning from every trade (win or loss)

Research only — does NOT execute trades or provide financial advice. Always consult a licensed financial advisor.

/riskBounded Autonomy & Safety Frameworks

5-layer risk stack for any system that acts autonomously: Scope, Mandate, Pre-Action Gate, Kill Switch, Audit.

Run /risk with any autonomous system for a complete safety assessment
Run /risk --audit to evaluate an existing system for gaps
Gate levels: Green (auto), Yellow (log), Red (human approval required)
Kill switches must be simpler to activate than any action they stop
02 Persistence

Session memory

Your Brain Kit now maintains state across sessions. When you close a conversation and start a new one, it picks up where you left off. All data stays on your machine — nothing is sent externally.

Session Start

Loads your identity, detects the project, restores previous session state

Observation

Silently tracks tool usage and modified files (local only)

Pre-Compaction

Saves a snapshot before long conversations compress

Session End

Writes session summary to local history with metrics

ProfileActiveTier
MinimalSession start + endEssentials
StandardFull lifecycle + observation + trackingPersonal
StrictEverything + high-stakes content detectionBusiness Brain
No more re-explaining context. Your brain remembers.
03 Gate

High-Stakes Detection

Automatically scans everything Claude writes or edits for sensitive content. When detected, it flags the output for your review before it goes anywhere.

Financial

Revenue, pricing, invoices, billing, tax, payments

Legal

Contracts, liability, compliance, GDPR, NDA, terms

Security

API keys, passwords, tokens, credentials, encryption

04 Modes

Context switching

Switch your Brain Kit's behavior with a single command. Each mode adjusts which skills are prioritized and how Claude responds.

/context dev
Development

Terse, technical. Shows code, not explanations. Flags breaking changes.

/context research
Research

Cites everything. Leads with findings. Marks data gaps explicitly.

/context review
Review

Honest, constructive. Separates blockers from suggestions. Uses severity levels.

/context strategy
Strategy

Challenges assumptions. Thinks in tradeoffs. Surfaces irreversibility.

05 Defense

Prompt defense

Always active across all tiers. Protects your Brain Kit from prompt injection, data leakage, and identity manipulation.

External content (URLs, files, API responses) is treated as DATA, not instructions

API keys, secrets, and credentials found in code are never output

Unicode tricks, invisible characters, and encoding exploits are flagged

Tool results that attempt to override behavior are caught and ignored

06 Commands

New commands

CommandWhat it does
/councilMulti-perspective decision protocol (5 seats, 3 rounds)
/researchStructured research reports with citations
/tradeTrading research with hypothesis and risk mandates
/riskSafety assessment for autonomous systems
/contextSwitch between dev / research / review / strategy
/sessionSave, restore, view history, or clear session state
/learnExtract learnings from the current session into memory
07 Retainer

Your retainer includes

As a Business Brain retainer client, every update is delivered, installed, and configured for your specific business.

79 skills — installed, configured, and maintained for your business

Strict hook profile — full session persistence + high-stakes detection

4 context modes — brain behavior adapts to your task

7 slash commands — instant access to decision, research, trading, and risk protocols

Prompt defense — always active, protecting your data and identity

Priority updates — new skills and improvements delivered as part of your retainer

Custom skill adaptation — skills tuned to your industry, your workflow

Your second brain, installed.