Both tools query overlapping venues (Polymarket and Manifold, plus Kalshi in one) for prediction market data, so an agent could be unsure which to call just to get odds. The descriptions do offer a meaningful distinction — one is a cross-venue arbitrage/spread scanner (paid), the other is a raw odds/probability lookup — which helps separate them, but the shared venue coverage creates real overlap.

Both names use snake_case, but they follow different conventions: kalshi_polymarket_arb is a venue-listing compound noun while prediction_markets is a generic noun with no verb or action cue. It is readable but there is no predictable verb_noun pattern to guide selection.

Two tools is thin for the stated scope of aggregating and scanning across three prediction markets. A single focused service could justify this, but the two tools carry a lot of combined responsibility (both multi-venue aggregation and arbitrage), leaving little granularity.

There is no way to list available markets/events, retrieve details or historical odds for a specific market, or select venues — the surface jumps straight to odds lookup and arbitrage scanning. For a prediction market aggregator this leaves significant gaps that could cause agent failures when it needs to discover what can be queried.