kraken-autonomy-levels
Progress from manual trading to full agent autonomy with controlled risk at each level.
- License 许可
- MIT
- 最近更新
- 2026-11-13
- Stars
- ★ 81
- Forks
- 17
01 前言 Front Matter
- name
- kraken-autonomy-levels
- version
- 1.0.0
- metadata
- { "openclaw": { "category": "finance" }, "requires": { "bins": [ "kraken" ] } }
- description
- Progress from manual trading to full agent autonomy with controlled risk at each level.
02 资源文件 3
- 许可证LICENSE1.0 KB
- 参考文档SOURCE.md171 B
- 参考文档TRUST.auto.yaml2.1 KB
03 SKILL.md 正文
kraken-autonomy-levels
Use this skill to help a trader move from hands-on CLI usage to fully autonomous agent trading in safe, incremental steps.
Level 1: Read-Only Agent
The agent reads market data and account state. No API key needed for public data; a query-only key for account data.
Kraken API key permissions: Query Funds, Query Open Orders & Trades
export KRAKEN_API_KEY="query-only-key"
export KRAKEN_API_SECRET="query-only-secret"
kraken ticker BTCUSD -o json 2>/dev/null
kraken balance -o json 2>/dev/null
kraken open-orders -o json 2>/dev/null
The agent can inform, alert, and recommend. It cannot place orders, cancel orders, or move funds.
Level 2: Paper Trading Agent
The agent tests strategies against live prices with zero risk. No API key needed.
kraken workspace create sandbox --capital 10000 --mode paper -o json 2>/dev/null
export KRAKEN_WORKSPACE=sandbox
kraken paper buy BTCUSD 0.01 -o json 2>/dev/null
kraken workspace status -o json 2>/dev/null
kraken workspace reset -o json --yes 2>/dev/null # dangerous: confirm with the user first
Run paper strategies repeatedly. Compare results across parameter changes. Only move to Level 3 after consistent paper performance.
Level 3: Supervised Trading Agent
The agent places real orders, but a human confirms each one. The API key has trading permissions, but the agent validates before executing and waits for human approval.
Kraken API key permissions: Query Funds, Create & Modify Orders, Cancel/Close Orders
unset KRAKEN_WORKSPACE # live orders run outside the paper workspace
export KRAKEN_API_KEY="trade-key"
export KRAKEN_API_SECRET="trade-secret"
# Step 1: Agent validates the order (no submission)
kraken order buy BTCUSD 0.001 --type limit --price 50000 --validate -o json 2>/dev/null
# Step 2: Agent presents the validated order to the human
# Step 3: Human approves or rejects
# Step 4: Agent executes only after approval
kraken order buy BTCUSD 0.001 --type limit --price 50000 -o json 2>/dev/null
Safety controls at this level:
- Always
--validatebefore executing - Always enable the dead man's switch for the session:
kraken order cancel-after 300 -o json - Check
open-ordersafter every trade to verify state
Level 4: Autonomous Trading Agent
The agent trades independently. No human in the loop. The trader trusts the agent and its strategy.
Kraken API key permissions: Create & Modify Orders, Cancel/Close Orders, Query Funds
export KRAKEN_API_KEY="full-trade-key"
export KRAKEN_API_SECRET="full-trade-secret"
The agent operates on its own schedule and passes --yes to the commands that prompt: cancel-all, withdrawals and transfers, earn allocations, workspace create --mode live, workspace reset, workspace promote, and recipe run --real-account. Placing an order never prompts at any level: the Level 3 approval is a step the agent takes itself.
Required safeguards at this level:
-
Restricted API key. Never use a key with withdrawal permissions for an autonomous agent. Create a key that can trade but cannot withdraw funds.
-
Dead man's switch. Start every session with
cancel-afterso all orders auto-cancel if the agent crashes or disconnects:kraken order cancel-after 600 -o json 2>/dev/nullRefresh periodically. If the agent stops refreshing, all orders cancel.
-
Position limits in agent code. The CLI does not enforce position size or frequency limits. Build these in your agent logic:
- Maximum order size per trade
- Maximum open positions
- Maximum trades per hour
- Pair allowlist (only trade specific pairs)
-
Error handling. Autonomous agents must handle every error category:
rate_limit: back off, reduce frequencynetwork: retry with backoff; ifverify_firstistrue, check state first so an order is never double-submittedauth: stop trading, alert the tradervalidation: log and skip, do not retry unchanged
-
Monitoring. Run a separate read-only agent (Level 1) that monitors positions and alerts the trader if thresholds are breached.
Level 5: Autonomous Agent with Fund Management
The agent can also move funds (deposits, withdrawals, transfers between wallets). This is the highest trust level.
Kraken API key permissions: All permissions including Withdraw Funds
This level is rare and high-risk. Only appropriate when:
- The agent manages a treasury or rebalances across exchanges
- The withdrawal addresses are pre-approved and whitelisted in Kraken's account settings
- Kraken's withdrawal address lock is enabled (prevents adding new addresses via API)
Choosing the Right Level
| Level | Agent can | Human involvement | Risk |
|---|---|---|---|
| 1 | Read data | None | Zero |
| 2 | Paper trade | None | Zero |
| 3 | Trade with approval | Confirms every order | Low |
| 4 | Trade autonomously | Monitors, intervenes if needed | Medium |
| 5 | Trade and move funds | Monitors | High |
Progression Rule
Move to the next level only after the agent demonstrates consistent behavior at the current level. Paper trading (Level 2) for at least a week before supervised trading (Level 3). Supervised trading for at least a week before autonomous (Level 4).
If you hit a mismatch between what you are trying to do and the CLI's interface or responses — including a mismatch between this skill and the installed CLI version's contract — feel free to submit feedback with kraken feedback.
04 同仓库更多 Skills
recipe-performance
Diagnose a specific MetaMask user flow's real-app performance and Sentry critical-user-flow telemetry with deterministic mm-harness recipes. Use for CUF boundaries, cold/warm/background-resume behavior, latency regressions, lifecycle tags, duplicate or blocking network requests, long tasks, layout/paint cost, screen redraws, React rerenders, or browser trace analysis where tests alone cannot prove the real app.
mexc-spot-websocket
MEXC Spot WebSocket workflows for public streams, protobuf handling, depth maintenance, listen keys, and private streams without exposing API keys.
mexc-spot-rest
MEXC Spot REST workflows for public data, signed account requests, orders, wallet/admin actions, and secure signing without exposing API keys.
mexc-futures-websocket
MEXC Futures WebSocket workflows for public streams, private account/order/position streams, ping/pong, subscriptions, and depth maintenance.
0xklen-version-and-refresh-sanctions-lists
暂无简介
verify-tests-cover-changed-lines
Use when a PR adds code that the existing suite may not exercise. Measures patch coverage — the share of newly changed lines hit by tests — before approving.