Arbor

Setup, intake, preflight, and launch-contract phase for open-source Arbor runs. Use when confirming a target project, metric, baseline, dev/test split, config/plugin settings, branch guard, session directory, or when translating a user goal into the precise contract consumed by the coordinator.

Public entrypoint for the Arbor skill suite. Use when a user wants to run an Arbor-style autonomous research or optimization workflow from a natural-language goal, including initial clarification of objective, target project, data, metric, evaluation, permissions, budget, run mode, and then automatic bootstrapping into arbor-agent-orchestrator and phase skills.

Coordinator phase for Arbor: persistent ReAct loop, Idea Tree state, INIT/OBSERVE/IDEATE/SELECT/DISPATCH/DECIDE protocol, tool mapping, cycle caps, and coordinator-only behavior. Use after setup/intake and before phase-specific ideation, executor, merge, search, or report skills.

Executor-dispatch phase for Arbor. Use when implementing an Idea Tree node through RunExecutor or RunExecutorParallel semantics: isolated git worktree, executor prompt construction, eval metadata injection, RunTraining policy, smoke/full evaluation, report parsing, artifact persistence, tree update, and insight propagation.

Strict IDEATE-stage skill for Arbor. Use immediately after TreeView(format="constraints") when drafting Idea Tree nodes, enforcing the idea_drafting and first_principles_probe behavior, depth-aware idea levels, four-line TreeAddNode hypotheses, conflict checks, and self-filtering against shallow tweaks.

Merge and evaluation discipline for Arbor. Use for TreeSetMeta metadata, B_dev/B_test separation, eval command templates, score parsing, GitMergeBranch behavior, protected paths, required outputs, metric_direction, trunk/test score updates, medal detection, and final evaluation before stopping.

Top-level controller for recreating the open-source AutoResearch workflow as a suite of skills. Use when the user asks to run, emulate, extract, validate, or refine Arbor/AutoResearch behavior, especially when a coordinator must load phase skills for setup, ideation, executors, merge evaluation, novelty search, plugins, resume, and reports.

Domain adaptation, human-in-the-loop, and budget policy phase for Arbor. Use when a run mentions plugins, plugin profiles, mle_kaggle, eval_contract, protected_paths, required_outputs, lifecycle hooks, convergence, budget_policy, RunTraining stages, AskUser, or interaction modes auto/direction/review/collaborative.

Resume, checkpoint, event, dashboard, finalization, and report phase for Arbor. Use when continuing interrupted sessions, handling running-node requeue, writing checkpoints/messages, consuming events.jsonl/run_stats, recovering best submissions, generating REPORT.md, or summarizing durable artifacts.

Related-work and novelty annotation phase for Arbor. Use after a node has been experimentally validated, especially before merge decisions, to emulate SearchIdeaContext/SearchIdeaContextParallel/SearchStatus, validated-node gating, background SearchAgent behavior, and structured prior-art summaries.

Deterministic helper layer for emulating Arbor tools in Codex or Claude Code when native TreeView, TreeAddNode, TreeSetMeta, RunExecutor, GitMergeBranch, or report tooling is unavailable. Use for local state management, eval score capture, executor prompt generation, merge checks, and skill forward tests.