harbor-cli - Harbor CLI evaluation and task management
Provides Harbor CLI commands for running evaluations, managing datasets and tasks, reviewing results, debugging trials, and exporting traces.
Tags
Updated: 2026-10-04Capabilities
Typical Inputs
Typical Outputs
What this skill does
- Run evaluation jobs
- Resume interrupted jobs
- Summarize job failures
- Run individual trials
- Summarize trial results
- List available datasets
- Download datasets
- Initialize benchmark adapters
- Review benchmark adapters
- Initialize tasks
- Assess task quality
- Launch task environments
- Debug failing tasks
- Browse evaluation results
- Run evaluation sweeps
- Export trace data
- Clean caches
Inputs
- Task directory paths
- Dataset names
- Job configuration files
- Trial configuration files
- Agent names
- Model identifiers
- Environment types
- Command-line options
- Registry URLs or paths
- Task names or paths
- Rubric files
- Job or trial paths
Outputs
- Evaluation results
- Trial results
- Summary files
- Downloaded task files
- Adapter scaffold files
- Task scaffold files
- Quality assessment results
- Web UI views
- Exported ATIF traces
- Cleaned Docker images and cache
Requirements
- Harbor CLI installation
- Docker or selected environment backend
- Model access for AI commands
