TerminalWorld is a benchmark dataset containing 1,530 terminal-based tasks for evaluating AI agents. It was created by EuniAI and last updated on May 22, 2026. The tasks were reverse-engineered from publicly available terminal recordings.
Use Cases
- Benchmarking AI agents on data processing tasks based on the described domain coverage
- Evaluating agent performance on system administration and networking tasks based on the described domain coverage
- Training or testing AI models on version control and container orchestration tasks based on the described domain coverage
- Assessing AI capabilities in debugging, testing, and environment setup based on the described domain coverage
Strengths
- Contains 1,530 distinct terminal-based tasks
- Covers multiple domains including data processing, system administration, networking, and security
Limitations
- Column-level documentation is absent; field semantics must be inferred after download
- Row count is unknown, which may limit suitability assessment
Provenance
- Source
- EuniAI
- Collection Method
- Reverse-engineered from publicly available terminal recordings
- Freshness
- Last updated 2026-05-22 12:21:57