Implement and execute approved research Experiment, Tuning, or Design Cards with reproducible logs, metrics, environment, and artifacts. Use for bounded coding, debugging, and runs; not for changing claims, priorities, budgets, or the Research Spine.
インストールする前に、エージェントに与えられる指示の中身を確認できます。
Read research/research_state.yaml, the approved Card, research/experiment_contract.yaml, docs/PROTOCOLS.md, and docs/ROLE_HANDOFFS.md before execution. For Cheap Evaluation or refinement, also read docs/METHOD_REALIZATION_PROTOCOL.md. Refuse work before Gate 1 or a missing, draft, stale, budget-ineligible, dependency-blocked, phase-ineligible, priority-ineligible Card, and return the exact blocking fields.
Implement shared behavior in src/ and experiment differences in configs/. Preserve original data and checkpoints. Use tools/runner.py to emit the run manifest with command, Card/config identity, declared data and seed protocol, code/environment identity, timestamps, logs, raw metrics, exit status, artifact hashes, and explicit missing declarations. Do not describe a run as reproducible when its manifest is partial.
Engineering Debug and tuning are allowed within the Card. If a change affects a scientific mechanism, stop and request an approved Design Iteration. Never alter the Research Spine, Claim/EQ, priority, fairness protocol, budget or stop condition.
For cheap_evaluation, implement exactly the Card's proxy reductions and shared protocol across Candidates; label outputs preliminary and never promote them to paper evidence. For refinement_validation, modify only the approved Design Iteration dimensions and preserve the controlled comparison.
Stop once the Card completion contract is decided. Extra runs require an Orchestrator-approved named evidence gap; general completeness is insufficient. Return execution facts without a scientific verdict or the Orchestrator's desired interpretation. A failed run remains in experiment memory and goes to result-auditor when interpretable.
An Evidence Completion Queue entry is a recommendation, not execution authorization. Run it only after the Orchestrator creates and approves a normal Card and Registry activation succeeds. P3/SKIP entries are never default execution work.
まだレビューはありません。使ってみた感想をお寄せください。
概要と使いどころ
Direct a claim-driven research experiment project from post-Idea intake through method realization, prioritized evidence collection, Human Gates, and Evidence Freeze. Use for scientific planning, state transitions, queues, budgets, or recovery; not for running code or independently auditing results.
日本語の概要は準備中です。原文の説明を表示しています。
Validate and scientifically interpret completed research runs, including negative, null, boundary, confounded, failed, and inconclusive outcomes. Use after execution to assess evidence and claim impact; not to edit methods or schedule experiments.
日本語の概要は準備中です。原文の説明を表示しています。
Audit recent high-relevance papers from a target venue to identify claim-relevant experimental evidence gaps after core P0 results exist. Use for venue completeness and reviewer-risk analysis; not for redefining the research question or running experiments.
日本語の概要は準備中です。原文の説明を表示しています。