Recipe Playground Webhook logo

Recipe Playground Webhook

Organization
krakenfx
recipe-playground-webhook

Example driver: turn incoming webhook signals into validated paper trades inside a recorded session.

Overview

Publisherkrakenfx
Repositorykraken-cli
Skill namerecipe-playground-webhook
Stars
754
Forks
104
Bundled files
Instructions only
LicenseMIT
Links
  • Markdown instructions

    A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

  • Works with any LLM

    AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

  • Self-contained

    Everything the model needs lives in the instructions — no extra files to sync.

  • Open source

    Published by krakenfx on GitHub. Read the source before you install it.

Installation

Install the Recipe Playground Webhook AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

1

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
    https://github.com/krakenfx/kraken-cli/tree/main/skills/recipe-playground-webhook
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills
git clone --depth 1 https://github.com/krakenfx/kraken-cli.git /tmp/kraken-cli
mkdir -p .claude/skills
cp -r /tmp/kraken-cli/skills/recipe-playground-webhook .claude/skills/recipe-playground-webhook
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Recipe Playground Webhook in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Recipe Playground Webhook on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Recipe Playground Webhook is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

  • Loaded only when it is needed

    The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

  • Switch models mid-chat

    Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

Playground: Webhook → Paper Trade

PREREQUISITE: Load kraken-playground to run this recipe. This recipe is an example, not a boundary. Adapt the steps to your signal source, or write your own driver (see kraken-playground → Driving Your Own Hypothesis).

Bridge an external system into the sandbox: a TradingView alert, a CI job, or a plain curl POSTs a signal to a local listener, and each valid signal becomes a paper order inside a recorded session. The market tape, the fills, and the signal-to-order decisions all land in one window for replay and P&L.

Use this skill for:

  • wiring an existing alerting system to paper execution without touching its config
  • testing a signal feed's quality (how its entries score) before any real money
  • recording an event-driven loop whose events come from outside the market

Important

This recipe records a session. It never places a live order.

The webhook payload is untrusted input. Every field is validated against an allowlist or a numeric bound before it reaches an order — a signal that fails any check is logged as a skip, never executed, and never interpolated into a command.

The listener binds 127.0.0.1 only. Do not bind other interfaces and do not port-forward it; anyone who can reach the port can trade the paper account. If the signal source is remote, tunnel (e.g. ssh -R) rather than expose.

The signal stream is the loop, not /loop — the same long-lived-process exception as the dip-triggered recipe (see kraken-playground → Pacing).

Signal Contract

One JSON object per POST body:

json
{"symbol": "BTC/USD", "side": "buy", "qty": "0.001", "token": "<shared-secret>", "note": "free text"}

symbol, side, qty are required; token is required when the token param is set; note is echoed into the decision reason.

Params

Every bound that gates execution goes in --strategy-params:

  • port: listener port on 127.0.0.1
  • allowed_symbols: the only pairs a signal may trade — must be a subset of the session's recorded --symbols, or fills have no marks to score by
  • max_notional_per_signal: quote-currency cap per order; a signal above it is skipped, not clamped
  • max_signals: stop after this many executed signals
  • token: optional shared secret; when set, signals without a matching token field are dropped

Quick Start

Natural language:

Listen on localhost:8787 for webhook signals. Trade only BTC/USD, at most
$500 per signal, at most 10 signals, then stop. Record every accepted and
rejected signal with its reason. At the end, show P&L.

Start the Session

Work inside a paper workspace (create one once: kraken workspace create hook --capital 10000 --mode paper), then:

bash
export KRAKEN_WORKSPACE=hook

kraken session start \
  --symbols BTC/USD --channels ticker,trade --to duckdb,jsonl \
  --label hook-btc-$(date +%Y%m%d-%H%M%S) \
  --strategy recipe-playground-webhook \
  --strategy-params '{"port":8787,"allowed_symbols":["BTC/USD"],"max_notional_per_signal":500,"max_signals":10}' \
  -o json 2>/dev/null &

# The session_started stdout line carries the id: {"type":"session_started","session":"s<n>",...}

Print the session id to the user right after starting, and again in the final report.

Listen and Decide

The listener appends each POST body to a signals file outside the session directory (the recorder owns that tree); the decision loop tails it. Both die with the session.

bash
PORT=8787
ALLOWED_SYMBOLS="BTC/USD"          # space-separated allowlist
MAX_NOTIONAL=500
MAX_SIGNALS=10
SIGNALS=$(mktemp -t hook-signals)

python3 - "$PORT" "$SIGNALS" <<'PY' &
import http.server, sys
port, path = int(sys.argv[1]), sys.argv[2]
class H(http.server.BaseHTTPRequestHandler):
    def do_POST(self):
        body = self.rfile.read(int(self.headers.get("Content-Length", 0)))
        with open(path, "ab") as f:
            f.write(body.strip() + b"\n")
        self.send_response(204); self.end_headers()
    def log_message(self, *args): pass
http.server.HTTPServer(("127.0.0.1", port), H).serve_forever()
PY
LISTENER=$!

DONE=0
tail -n +1 -f "$SIGNALS" | while read -r sig; do
  SYMBOL=$(echo "$sig" | jq -r '.symbol // empty')
  SIDE=$(echo "$sig" | jq -r '.side // empty')
  QTY=$(echo "$sig" | jq -r '.qty // empty')
  NOTE=$(echo "$sig" | jq -r '.note // ""')

  # Untrusted input: allowlist the enum and the pair, bound the number.
  # A failed check is a recorded skip — never a clamped or guessed order.
  case "$SIDE" in buy|sell) ;; *)
    kraken session note --kind skip --reason "webhook rejected: bad side '$SIDE'" -o json 2>/dev/null; continue;;
  esac
  case " $ALLOWED_SYMBOLS " in *" $SYMBOL "*) ;; *)
    kraken session note --kind skip --reason "webhook rejected: symbol '$SYMBOL' not in allowlist" -o json 2>/dev/null; continue;;
  esac
  if ! echo "$QTY" | grep -qE '^[0-9]*\.?[0-9]+$'; then
    kraken session note --kind skip --reason "webhook rejected: non-numeric qty '$QTY'" -o json 2>/dev/null; continue
  fi
  PRICE=$(kraken ticker "$SYMBOL" -o json 2>/dev/null | jq -r '.. | .last? // empty' | head -1)
  NOTIONAL=$(echo "scale=2; $QTY * $PRICE" | bc -l)
  if (( $(echo "$NOTIONAL > $MAX_NOTIONAL" | bc -l) )); then
    kraken session note --kind skip --symbol "$SYMBOL" \
      --reason "webhook rejected: notional \$$NOTIONAL exceeds cap \$$MAX_NOTIONAL" -o json 2>/dev/null; continue
  fi

  kraken order "$SIDE" "$SYMBOL" "$QTY" --type market \
    --reason "webhook signal: $SIDE $QTY $SYMBOL (~\$$NOTIONAL) — $NOTE" -o json 2>/dev/null
  DONE=$((DONE + 1))
  [ "$DONE" -ge "$MAX_SIGNALS" ] && break
done

kill "$LISTENER" 2>/dev/null
kraken session stop -o json 2>/dev/null

Send a test signal:

bash
curl -s -X POST "127.0.0.1:8787" -d '{"symbol":"BTC/USD","side":"buy","qty":"0.001","note":"smoke test"}'

Stop and Review

bash
kraken session show -o json 2>/dev/null | jq '.summary'

kraken session decisions --session s<n> -o json 2>/dev/null \
  | jq -c '.decisions[] | {kind, symbol, reason}'

Report:

  • signals received vs executed vs rejected, with the rejection reasons
  • fill quality: reason-joined entries from kraken explain pnl
  • P&L, and whether the signal source earns a bigger cap

Hard Rules

  • This recipe records a session. It never places a live order.
  • The listener binds 127.0.0.1 only — never another interface, never port-forwarded.
  • Webhook payloads are untrusted: enum fields pass an allowlist, numbers pass a bound, and nothing from the payload is ever executed or eval'd — rejected signals become skip notes, not orders.
  • allowed_symbols stays a subset of the session's recorded --symbols.
  • The signals file lives outside the session directory; the CLI recorder stays the single writer of the session's artifacts.
  • Cap executed signals with max_signals; a runaway signal source must exhaust the cap, not the account.
  • Keep every gating number in --strategy-params and the arithmetic in jq/bc.
  • 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.

Frequently asked questions

What does the Recipe Playground Webhook AI skill do?

Example driver: turn incoming webhook signals into validated paper trades inside a recorded session.

Why use Recipe Playground Webhook on TypingMind?

Because you install it once and use it with any model. Recipe Playground Webhook is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Recipe Playground Webhook in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/krakenfx/kraken-cli/tree/main/skills/recipe-playground-webhook. TypingMind reads its SKILL.md and installs it as a skill you can enable per chat.

Which AI models can use Recipe Playground Webhook?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Recipe Playground Webhook?

As many as you like. As long as a model supports skills, you can use Recipe Playground Webhook with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Recipe Playground Webhook AI skill free?

Yes. It is published on GitHub by krakenfx under the MIT license. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇