Public leaderboard

Public assessment

jordanlyall/wc26-mcp (wc26-mcp)

wc26-mcp · v0.3.1 · scanned

What changed in the harness

Selection accuracy 98→98, token cost down 2%, unconfirmed writes 100%→100%.

Category breakdown

Where the score comes from.

Earned points across the four signals Gradable measures. Safety and Legibility are scored out of 30; Economics and Discoverability are scored out of 20.

01Safety

30.0 / 30

30.0 out of 30
02Legibility

26.3 / 30

26.3 out of 30
03Economics

18.5 / 20

18.5 out of 20
04Discoverability

13.8 / 20

13.8 out of 20

Highest-impact fix

Estimated gain +6 points

Make target tools discoverable on the first call

Clarify tool names, decision boundaries, and required argument schemas so an agent can choose and construct the target call without exploratory steps.

Description evidence

Defects and rewrites.

1 defect found across the exposed tool descriptions. Suggested rewrites make purpose, inputs, boundaries, and returns easier for an agent to understand.

Tool Defect types Suggested rewrite
get_injuries
no_return_description
Track key player availability for the FIFA World Cup 2026. Returns a list of star players across all 48 teams with their current injury status (out, doubtful, recovering, or fit). Filter by team ID or FIFA code (e.g. 'usa', 'BRA') or by status; omit both filters to see all tracked players at once.

Selection evidence

Confusable tool pairs.

12 pairs where similar names or overlapping descriptions may send an agent toward the wrong tool.

Tool A Tool B Confidence Why they collide
get_teams get_team_profile medium A generic task like 'tell me about the USA team' could route to either: get_teams returns rankings/details while get_team_profile returns the full profile, and both use the same team/group concepts.
get_venues get_nearby_venues medium 'What venues are around New York' or 'venues close to MetLife' could trigger get_venues (city filter) instead of get_nearby_venues (proximity), since the names differ only by 'nearby'.
get_matches get_groups medium 'What are the matches in group B?' fits both: get_matches has a group filter and get_groups returns each group's match schedule.
get_matches get_bracket medium 'Show me the round of 16 / knockout matches' is served by both get_matches (round filter) and get_bracket (knockout rounds with venues and dates).
get_groups get_standings medium 'Show me group A info/standings' — both accept a group letter; get_standings returns power rankings while get_groups returns teams and schedule, so group-level queries are genuinely ambiguous.
get_odds get_standings high Both cover group-by-group predictions derived from betting odds; get_standings explicitly cites betting odds and predictions, so 'who is expected to win group A' has no clear single winner.
get_news get_injuries medium 'Any injury updates for Brazil?' could route to get_news, which has an explicit 'injury' category, instead of the dedicated get_injuries tracker.
get_historical_matchups compare_teams high Both accept team_a/team_b and compare_teams explicitly advertises head-to-head history, so 'how has Argentina vs Brazil gone historically' could land on compare_teams rather than the dedicated get_historical_matchups.
get_matches get_teams low 'Which teams are playing in group A' fits both group filters — get_teams returns team details, get_matches returns team names per match — though primary intent usually disambiguates.
get_groups get_schedule low 'Show the schedule for group C' — get_groups embeds each group's match schedule while get_schedule covers the whole tournament, so schedule-producing queries can hit either.
get_city_guide get_fan_zones low 'What is there to do in Dallas during the Cup' could route to the city travel guide (things to do) or fan zones (activities/amenities), as both are host-city leisure resources.
get_matches get_venues low 'Which venues host matches in Mexico' — get_matches returns venue details per match and get_venues returns venues with notable events, so venue-focused match queries are mildly ambiguous.

Compare the field

One score is useful.
The evidence makes it actionable.

Back to the leaderboard