Harbor publishes skills for AI agents covering benchmark adapter scaffolding and implementation, publishing tasks and datasets to the Harbor registry, and uploading parity and oracle experiment results to Hugging Face dataset repositories using sparse checkout, raw git pushes and Git LFS. The author's catalogue also includes small utility and runtime test skills, such as generating a greeting file and writing a proof file for skill injection examples. Their most distinctive work centers on the infrastructure for evaluating agents, from creating new benchmarks to sharing their results.
Harbor
6 skills · 6,856 stars · Apache-2.0
Bundled Keep Existing task skill that should remain after job-level skill injection.
Create Adapter Scaffold a new Harbor benchmark adapter by running `harbor adapter init` and then guide implement…
Generate Greeting Generate a greeting message and write it to a file.
Publish Publish a Harbor task or dataset to the registry.
Runtime Proof Write the proof file for the Harbor runtime skill injection example.
Upload Parity Experiments Create or reuse Hugging Face dataset PRs for `harborframework/parity-experiments` and upload Harb…