Terminal-Lego-15k: Docker-Verified Terminal Agent Tasks from StackOverflow

Terminal-Lego-15k is a large-scale collection of Docker-verified, Terminal-Bench-style agentic tasks built from real StackOverflow technical issues. The dataset was created by SWE-Lego through a pipeline that filters questions, converts them via cascaded LLM generation, and retains tasks only after Docker round-trip verification. It was last updated on June 4, 2026.

Use Cases

• Train terminal agents to execute commands based on real-world technical problems.
• Benchmark the performance of AI agents on Docker-verified command-line tasks.
• Study the translation of StackOverflow questions into executable, verifiable workflows.

Strengths

• Contains 15,000 tasks, indicating a substantial scale for training and evaluation.
• Tasks are derived from real StackOverflow technical issues, providing practical relevance.
• Each task undergoes Docker round-trip verification, suggesting a focus on functional correctness.

Limitations

• Row count is unknown, which may limit suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

StackOverflow technical issues.

Collection Method

Filtered and converted via a cascaded LLM generation pipeline, with Docker round-trip verification.

Freshness

Last updated 2026-06-04 09:12:02; freshness should be verified.