Prediction Markets Monitoring

Prediction Markets Monitoring

Use this skill when the user wants to track one or more Kalshi prediction markets over time, check what changed, refresh a thesis, compare a watchlist of related contracts, or inspect model-vs-market probability deltas.

Workflow

1. Collect Kalshi URLs or event tickers from the user. Pass full Kalshi URLs directly to Octagon tools; do not use Bash, curl, or direct Kalshi API calls for preprocessing.
2. Use octagon_prediction_market_research with cacheMode: "cache" first for fast/free cached context. Treat cached research as the prior thesis, not necessarily the current market state.
3. Use octagon_prediction_market_history second with a bounded limit, optional days, and excludeEmptyModel: true when the user wants cleaner recent model data. Use excludeEmptyModel: false when the user mainly wants current market-implied odds, because model fields may be empty before a report exists.
4. Compare cached report freshness against the latest history snapshot:
   • Less than 24 hours old: usually safe to use as current context unless the market moved sharply.
   • More than 24 hours old on an active/volatile market: recommend refresh before making a model-edge claim.
   • More than 72 hours old on a slower market: recommend refresh if the user needs a live thesis.
   • Around 5 days old or older: strongly recommend refresh and label the cached report as stale prior context.
5. If no cached report exists, avoid claiming an Octagon model probability or edge. Use history for market-implied probabilities and say refresh is required to create the first full Octagon report/model view.
6. Use octagon_prediction_market_compare for multiple tickers and include history when the user asks what changed.
7. Use octagon_prediction_market_research with cacheMode: "refresh" only when the user asks for a fresh catalyst/probability update, no report exists and they want one created, or the cached report is stale. Fresh reports are available to Octagon Plus, Pro, and Enterprise users and cost 3 credits.
8. After a successful refresh, immediately re-run cacheMode: "cache" and recent history. The ideal final answer compares the refreshed report against recent price/probability movement.
9. If a default or refresh report request times out, do not stop. Treat fallback output as partial success: cached report status and market history can still answer what the market implies, even when a fresh Octagon model report did not complete.
10. Keep the output focused on probability deltas, edge changes, confidence, liquidity, new drivers, report age, and whether the prior thesis still holds.

Avoid persistent watchlist assumptions. This v1 plugin provides analysis tools but does not create background jobs or store watchlists.

Output Pattern

Use this structure for monitoring answers:

1. Current market snapshot from history.
2. Cached Octagon report status: fresh, stale, or missing.
3. What moved: probability, volume, confidence, or edge changes when available.
4. Recommendation: use cache, use refresh, or watch without refreshing.
5. Caveat: say when model probability/edge is unavailable because no report exists.
6. Provenance: say whether the answer comes from a fresh report, cached report, or market-history fallback.