Kalshi Hybrid Signal Trader

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

• find kalshi trades
• rank kalshi opportunities
• explain a kalshi contract with macro support

On Invoke

Immediately run the default Kalshi scan without asking. Do not present a menu. Execute:

cd ~/.config/seren/skills/kalshi-hybrid-signal-trader && python3 scripts/agent.py --config config.json --mode scan

If the user explicitly requests paper or live, run that mode instead. After results are displayed, present the next steps.

Output Contract

This skill is self-contained. Use these local references:

• references/output-contract.md
• references/desktop-summary-template.md

Human-facing output must be rendered first as a mini research note. Structured agent payload comes immediately after it using kalshi-shared-v1.

Workflow Summary

1. fetch_kalshi_markets uses connector.kalshi_oracle.get
2. fetch_gap_signals uses connector.macro_gap.get
3. fetch_coil_signals uses connector.macro_coil.get
4. assess_signal_health uses transform.assess_kalshi_signal_health
5. score_market_candidates uses transform.score_kalshi_hybrid_candidates
6. rank_candidates uses transform.rank_kalshi_candidates
7. create_trade_intents uses transform.create_kalshi_trade_intents
8. persist_run uses connector.storage.post
9. persist_candidates uses connector.storage.post
10. persist_trade_decisions uses connector.storage.post
11. render_summary uses transform.render_kalshi_hybrid_report

Execution Modes

• scan (default): return ranked dry-run trade intents plus watchlist and health summary.
• paper: same selection logic, but mark the selected trades as paper-only records suitable for persistence or downstream automation.
• live: requires explicit --yes-live and still fails closed if signal health is ambiguous. This runtime emits trade intent output only; it does not submit broker orders.

Product Behavior

The trader combines:

• Kalshi oracle divergence
• gap support
• coil support
• freshness and health interpretation
• plain-language rationale

If gap or coil is ambiguous, downgrade the candidate to watchlist-only and explain the downgrade explicitly.

Human-Facing Output

For every surfaced contract, explain:

• what the Kalshi contract is
• what divergence or consensus break was observed
• what gap saw
• what coil saw
• whether it is tradable or watchlist-only
• what key risk could make the view wrong
• whether freshness or health caveats are present

Shared Output Fields

This skill must return:

• run_status
• mode
• generated_at
• signal_health
• market_candidates
• selected_trades
• watchlist
• blocked_reasons
• rationale
• risk_note
• freshness
• desktop_summary
• audit

Trade Execution Contract

When the user gives a direct exit instruction such as sell, close, exit, unwind, or flatten, treat it as an instruction to cancel or clear any planned Kalshi trade intents immediately. Ask only the minimum clarifying question if the target contract is ambiguous.

Pre-Trade Checklist

Before marking a contract as live-eligible:

1. Fetch the current Kalshi oracle payload and confirm the contract is still active.
2. Check gap and coil freshness and interpret empty responses explicitly.
3. Verify required publishers are reachable and SEREN_API_KEY is loaded.
4. If signal health is ambiguous, fail closed and downgrade to watchlist-only.
5. Return a customer-readable explanation of the downgrade, not just an empty trade list.

Dependency Validation

The runtime must verify required credentials and publishers before continuing:

• SEREN_API_KEY
• seren-polymarket-intelligence
• serendb

If anything is missing, stop with a remediation message instead of guessing or attempting live execution.

Persistence

Persist at minimum:

• scan runs
• ranked Kalshi candidates
• selected trade intents
• watchlist entries
• signal health snapshots
• rationale payloads