Create Hive task repositories with program.md and eval scripts.
npx skills add https://github.com/rllm-org/hive --skill hive-create-task
Please help me install this Agent Skill. Skill: hive-create-task Source: https://github.com/rllm-org/hive/tree/main/claude-plugin/skills/hive-create-task Command: npx skills add https://github.com/rllm-org/hive --skill hive-create-task
Helps users turn an idea into a fully scaffolded, tested, and publishable Hive task repository by guiding them through problem definition, eval design, constraints, repo scaffolding, baseline testing, and upload workflows.
Ask the skill to design and scaffold a new hive task for your idea, including eval design and a starter repo.
High-intent search queries and answers about installing and using this skill.
You create a runnable task repository for an ML training benchmark by using an interactive wizard to define problem scope, scaffold program.md and artifacts, and generate a git-ready repo with a parseable eval script.
For agentic benchmark tasks, the eval script must print a standard score block, specify artifact and prepare scripts, and generate parseable output to ensure the repository passes verification and upload checks.
Yes, you can scaffold a coding challenge repository without manual setup by using a conversational wizard that automatically templates program.md, eval/eval.sh, requirements.txt, prepare.sh, and starter artifact files for immediate testing.
You package a prompt engineering task for upload and verification by generating a complete repository structure with program.md, a parseable eval script, and git-ready artifacts through a step-by-step interactive scaffolding process.
The best way to diagnose eval script failures when designing reproducible tasks is to run prepare and eval loops iteratively, allowing the wizard to suggest adjustments to baselines or eval scripts until the repository passes verification.
You do not need to pre-format dataset requirements before scaffolding a Hive task repository, as the interactive wizard guides you through dataset selection and task metadata clarification step-by-step during the creation process.
yu13130122297Identify uncertainties in research framing, datasets, metrics, and baselines.
Vasallo94Run and debug Python test suites with pytest and uv.
greencmMerge pull requests, run CI tests, and deploy to production with canary checks.
EntityProcessOrchestrate AgentV evaluations by simulating LLM responses without API keys.
