High-Throughput Paired Basis Maker

For Claude: How to Use This Skill

Skill instructions are preloaded in context when this skill is active. Do not perform filesystem searches or tool-driven exploration to rediscover them; use the guidance below directly.

When to Use

• trade relative-value dislocations between logically linked Polymarket contracts
• enforce backtest-first validation before generating paired trade intents
• run a dry-run-first workflow for hedged pair execution

On Invoke

Immediately run the default paired-market backtest without asking. Do not present a menu of modes. Execute:

cd ~/.config/seren/skills/high-throughput-paired-basis-maker && source .venv/bin/activate && python3 scripts/agent.py --config config.json

Display the full backtest results to the user. Only after results are displayed, present available next steps (trade mode). If the user explicitly requests a specific mode in their invocation message, run that mode instead.

Backtest Period

• Default: 270 days
• Allowed range: 90 to 540 days
• Why this range: basis relationships need enough time to observe repeated widening/convergence cycles, but should still emphasize current structural behavior.

Workflow Summary

1. load_backtest_pairs pulls live market histories from the Seren Polymarket Publisher (Gamma + CLOB proxied), attaches per-leg order-book snapshots, builds pairs from the active market universe, and timestamp-aligns each pair.
2. simulate_basis_reversion runs an event-driven stateful replay with carried cash and inventory across both legs, order-book-aware fills, and pessimistic spread-decay.
3. summarize_backtest reports total return, annualized return, Sharpe-like score, max drawdown, hit rate, quoted/fill counts, order-book mode coverage, telemetry counts, and pair-level contributions.
4. sample_gate fails backtest if events < backtest.min_events (default 200).
5. backtest_gate blocks trade mode by default if backtest return is non-positive.
6. emit_pair_trades outputs two-leg trade intents (primary + pair) with risk caps.

Execution Modes

• backtest (default): paired historical simulation only.
• trade: always runs backtest first, then emits paired trade intents if gate passes.

Live execution requires both:

• execution.live_mode=true in config
• --yes-live on the CLI
• POLY_PRIVATE_KEY (or WALLET_PRIVATE_KEY) plus POLY_API_KEY / POLY_PASSPHRASE / POLY_SECRET

Trade Execution Contract

When the user gives a direct exit instruction (sell, close, exit, unwind, flatten), execute the exit path immediately. Do not editorialize or argue against recovering remaining funds. If the user request is ambiguous, ask only the minimum clarifying question needed to identify the positions to exit.

CLOB Exit Rules

• py-clob-client via DirectClobTrader is the canonical live execution path.
• For immediate sells, use a marketable limit priced at the market minimum tick from the live order book. Do not hardcode $0.001; use the current tick_size.
• Never place a passive sell above the best bid when the user asked for an immediate exit.
• Estimate recovery by sweeping visible bid levels (price x size) across the full book, not just the best bid.
• If visible bid depth cannot cover the full exit size, report the partial-depth estimate and remaining unfilled size.

Pre-Trade Checklist (Mandatory)

Before any live buy, sell, or unwind:

1. Fetch the live order book for every token involved.
2. Snap prices to the market tick_size and compute visible-book recovery or cost across all levels.
3. Verify the current environment can legally and technically reach the Polymarket CLOB API. If access is blocked, stop and report the restriction; do not suggest bypasses.
4. Verify py-clob-client is installed and POLY_PRIVATE_KEY or WALLET_PRIVATE_KEY, POLY_API_KEY, POLY_PASSPHRASE, and POLY_SECRET are loaded.
5. If any dependency check fails, fail closed with a concrete remediation message.

Emergency Exit

Immediately liquidate held inventory with:

python3 scripts/agent.py --config config.json --unwind-all --yes-live

The unwind path cancels open orders first, then submits marketable min-tick sells and reports visible-book exit estimates.

Runtime Files

• scripts/agent.py - basis backtest + paired trade-intent runtime
• scripts/setup_cron.py - create/update the skill-local seren-cron local-pull runner and job
• scripts/run_local_pull_runner.py - poll seren-cron and execute due local jobs on this machine
• config.example.json - strategy parameters and live backtest defaults
• .env.example - environment template for API credentials
• requirements.txt - installs py-clob-client for live order signing/submission

API Key Setup

Before running this skill, check for an existing Seren API key in this order:

1. Seren Desktop auth — if the skill is running inside Seren Desktop, the runtime injects API_KEY automatically. Check: echo $API_KEY. If set, no further action is needed.
2. Existing .env file — check if SEREN_API_KEY is already set in the skill's .env file. If set, no further action is needed.
3. Shell environment — check if SEREN_API_KEY is exported in the current shell. If set, no further action is needed.

Only if none of the above are set, register a new agent account:

curl -sS -X POST "https://api.serendb.com/auth/agent" \
  -H "Content-Type: application/json" \
  -d '{"name":"high-throughput-paired-basis-maker"}'

Extract the API key from the response at .data.agent.api_key — this key is shown only once. Write it to the skill's .env file:

SEREN_API_KEY=<the-returned-key>

Verify:

curl -sS "https://api.serendb.com/auth/me" \
  -H "Authorization: Bearer $SEREN_API_KEY"

Do not create a new account if a key already exists. Creating a duplicate account results in a $0-balance key that overrides the user's funded account.

Reference: https://docs.serendb.com/skills.md

Quick Start

cd ~/.config/seren/skills/polymarket-high-throughput-paired-basis-maker
python3 -m venv .venv
source .venv/bin/activate
python -m pip install --upgrade pip
python -m pip install -r requirements.txt
cp .env.example .env
cp config.example.json config.json
python3 scripts/agent.py --config config.json

If you are already running inside Seren Desktop, the runtime can use injected auth automatically.

Live market data only. Always leave "markets": [] and "state": {"leg_exposure": {}} empty in your config.json. The skill discovers and fetches live Polymarket pairs automatically via backtest.gamma_markets_url. Never add placeholder market IDs — they do not exist on Polymarket and will cause the backtest to fail with "No markets with sufficient history".

Run Trade Mode (Backtest-First)

python3 scripts/agent.py --config config.json --run-type trade

Optional Fixture Replay

python3 scripts/agent.py --config config.json --backtest-file tests/fixtures/backtest_pairs.json

Set backtest.telemetry_path to capture JSONL replay telemetry for each decision step.

Seren Predictions Intelligence

After a backtest completes, the output will suggest enabling Seren Predictions if it is not already active. This optional feature uses computed pair-specific endpoints to:

• Validate pair correlation strength with cross-platform data
• Rank pairs by basis deviation sigma for better entry signals
• Filter for pairs where cross-platform data confirms mean-reversion potential

Endpoints used (computed, not consensus):

• GET /api/polymarket/pairs/suggested ($0.10) — suggested pairs ranked by basis deviation
• GET /api/polymarket/correlations ($0.10) — pair correlation and basis spread statistics

To enable, set predictions_enabled: true in the backtest section of your config.json. Estimated cost: ~$0.20 SerenBucks per backtest run.