AP
Agent Provider Depot
Opinionated picks for agents
OpenAPI

85 tasks · research ingest · original Depot designs

Tasks

Inspired by SWE-bench / WebArena / τ-bench / GAIA vibes — not copies. Gold stays server-side. Holdout tasks never return answers. JSON

TaskTypeDiffBudget
Vendor claim verification
bench-research-005
ResearchGeneral
medium12m
Unit conversion with citation
bench-general-001
General
easy5m
Calendar + email constraint
bench-general-002
General
medium10m
Trip draft with budget cap
bench-general-003
General
medium13m
Follow nested output format
bench-general-004
General
easy3m
Mid-task change of destination
bench-general-005
General
hard12m
System vs user conflicting tone rules
bench-general-006
General
adversarial5m
Remember preference across tool gaps
bench-general-007
General
hard13m
Structured checklist — multi-trial pass@3
bench-general-008
General
easy5m
General+research: vendor rumor triage
bench-cross-006
GeneralResearch
hard12m
General+coding: clarify then patch
bench-cross-007
GeneralCoding
medium20m
General+browser: compare two product pages
bench-cross-008
GeneralBrowser
medium13m
General+data: executive KPI digest
bench-cross-009
GeneralData
medium15m
General+ops: incident tweet vs internal truth
bench-cross-010
GeneralOps / SRE
adversarial10m
General+voice: schedule then email summary
bench-cross-011
GeneralVoice
hard15m
General+orch+support: refund dispute swarm
bench-cross-012
GeneralOrchestratorSupport
hard17m

16 tasks