SkillFlow Test Tasks: 166 Runnable Tasks for Agent Benchmarking

SkillFlow Test Tasks is a repository of 166 runnable tasks used in the SkillFlow benchmark for evaluating autonomous agents. The tasks are organized into 20 workflow families spanning five domains, including finance, healthcare, and governance. The dataset was created by author zhang-ziao and was last updated on 2026-04-21.

Use Cases

• Benchmarking lifelong skill discovery in agents based on the 166 runnable tasks
• Evaluating skill revision and adaptation based on the 20 workflow families
• Testing cross-task procedural transfer in autonomous agents across the five broad domains

Strengths

• Contains 166 distinct runnable tasks for agent evaluation
• Organized into 20 workflow families across five broad domains

Limitations

• Column-level documentation is absent; field semantics must be inferred after download
• Row count is unknown, which may limit suitability assessment

Provenance

Source

huggingface

Collection Method

Likely created as a benchmark for autonomous agent research.

Freshness

Last updated 2026-04-21 05:30:01; freshness should be verified