adrianosousa/oikos-strategies

Oikos Strategy Management

Strategies are declarative .md files that guide your financial behavior. They define portfolio targets, DeFi triggers, swarm engagement rules, and risk limits — all operating within the hard PolicyEngine constraints.

Strategies are intent, not code. You interpret them on each decision. Humans can read and edit them. They're tradeable on the swarm.

MCP Tools

| Tool | Purpose | |------|---------| | get_active_strategies | Read all active strategies before any financial decision | | save_strategy | Create or update a strategy file | | toggle_strategy | Enable/disable a strategy by filename |

Strategy File Format

Every strategy is a markdown file with YAML frontmatter:

---
enabled: true
source: agent          # "human", "agent", or "purchased"
version: 1.0
created_at: 2026-03-20T18:00:00Z
expires_at: 2026-03-21T18:00:00Z   # Optional: auto-expire for agent-authored
requires_approval: false             # Optional: require human approval to activate
confidence: 0.85                     # Optional: agent's confidence in this strategy
tags: [defi, yield, rebalancing]
---

# Strategy Name

## Section (Portfolio Targets, Trigger Conditions, Actions, Risk Limits, etc.)
- Declarative rules in plain language

Decision Flow

1. Before any financial decision, call get_active_strategies
2. Parse enabled strategies and factor them into your reasoning
3. If a strategy suggests an action (rebalance, swap, yield deposit), verify it passes PolicyEngine via simulate_proposal
4. Execute only if both strategy intent and policy approval align
5. If you author a new strategy, set source: agent and expires_at (24h default)

Strategy Authoring Rules

When you write a strategy:

• Set enabled: false initially — confirm with human before activating
• Always include expires_at (time-boxed by default, renewable)
• Use source: agent to distinguish from human-authored strategies
• Include a confidence score for the strategy itself
• Write declarative intent, not imperative instructions
• Reference PolicyEngine limits — your strategy operates WITHIN them, never overrides

Swarm Strategy Trading

Strategies can be bought and sold on the swarm marketplace:

• Selling: Use swarm_announce with category seller, include strategy description
• Delivering: Use swarm_deliver_result to send the strategy file after payment
• Purchasing: When you buy a strategy, save it with source: purchased and enabled: false
• Quarantine: Purchased strategies should be reviewed before enabling

Example: Yield-Optimizing DeFi Agent

---
enabled: true
source: agent
version: 1.0
created_at: 2026-03-20T18:00:00Z
expires_at: 2026-03-21T18:00:00Z
confidence: 0.85
tags: [defi, yield, rebalancing, autonomous]
---

# Yield-Optimizing DeFi Agent Strategy

## Portfolio Targets
- 50% USDT — working capital + Aave yield
- 30% XAUT — store of value
- 20% ETH — gas + exposure

## Trigger Conditions
- Any asset drifts >10% from target -> rebalance
- Idle USDT >300 -> deposit excess to Aave
- Aave APY <3% -> withdraw and hold
- ETH drops >8% in 24h -> pause non-essential swaps

## Risk Limits
- Never swap >20% of portfolio in one day
- Never borrow (no leverage)
- Never bridge without human approval
- Portfolio drops >15% in 24h -> halt all actions, alert human