Coordinate iterative artifact improvement against measurable objectives with hypothesis trees and held-out gating.
npx skills add https://github.com/dralkh/seerai --skill arbor
Please help me install this Agent Skill. Skill: arbor Source: https://github.com/dralkh/seerai/tree/main/skills/arbor Command: npx skills add https://github.com/dralkh/seerai --skill arbor
It coordinates long-horizon improvement of a concrete artifact against a measurable objective, so you can run many experiments without losing track of what was learned.
Ask Arbor to improve your current artifact against a named dev evaluator and held-out test evaluator, then let it run iterative experiments, record evidence, and merge only the changes that generalize.
High-intent search queries and answers about installing and using this skill.
Iterative experiment tracking uses a hypothesis tree to organize competing directions and isolated worktree execution to test one hypothesis at a time, recording evidence to prevent knowledge loss. It coordinates long-horizon improvement of a concrete artifact against a measurable objective.
Dev/test separation uses a fast development evaluator during the search phase and admits changes only after held-out confirmation. This merge gating prevents overfitting to evaluator metrics and ensures changes generalize before merging.
Isolated worktree execution tests one hypothesis per worktree to avoid cross-contamination between concurrent changes. Each worktree runs independently against a dev evaluator, recording evidence before merging the successful artifact updates.
Yes, prompt refinement and agent harness tuning are core use cases alongside model training recipes, data pipelines, and benchmark submissions. Any task requiring iterative improvement of a concrete artifact against a measurable objective applies.
Yes, a hypothesis tree is required to organize competing directions, refinements, and negative constraints before executing isolated runs. It structures the search space so that evidence propagation and held-out merge gating can operate effectively.
Held-out merge gating prevents overfitting by only merging changes that pass a separate test evaluator after succeeding on the fast dev evaluator. This ensures iterative artifact improvements generalize rather than just fitting the development metric.
jasrajtulsiCoordinate autonomous optimization of an artifact against a measurable objective.
bdsqqqRecreate the current branch with clean, narrative commit history.
soo-kate-yeonCreate isolated Git worktrees for parallel SPEC development with MoAI-ADK integration.
SignmanalTrack ML experiment metrics, hyperparameters, and model versions with W&B.
