菜单
Route AI research workspace work across the evidence graph. Use for broad research tasks, choosing which research sub-skill applies, or updating the dashboard after experiment, claims, synthesis, or reference work.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Turn a research agenda into one cold-readable `/goal` prompt for autonomous AI research loops, with evidence-graph notebook rules, experiment cadence, verification, scope edges, and a finite, batch, or continuous completion contract.
Maintain the research workspace belief layer: open questions, claims, decisions, confidence levels, evidence links, and revisit conditions.
Rank candidate experiments, decide what not to run next, maintain the factory queue, and choose one decisive next research goal before autonomous loops spend compute.
Create, run, and record AI research experiments with manifests, run records, artifact paths, metrics, and experiment index updates.
Compress an AI research experiment batch or branch into lessons, updated claims/questions, and a continue/stop/pivot decision.
Remote CUDA development workflow for the local NVIDIA DGX Spark at spark-f0a3.local. Use when Codex needs to probe the DGX Spark, choose a developer experience, mirror a research workspace with rsync, run CUDA/PyTorch/NAC/Le-WM training remotely while the agent stays on the Mac, use tmux/screen, manage TensorBoard tunnels, sync artifacts back, or decide between direct SSH, Docker, NVIDIA AI Workbench, and native Python setup.
| name | research-workspace |
| description | Route AI research workspace work across the evidence graph. Use for broad research tasks, choosing which research sub-skill applies, or updating the dashboard after experiment, claims, synthesis, or reference work. |
This is the router skill. Use it first for broad research work, then load the smallest sub-skill that matches the task:
research-experiment: creating experiments, running commands, recording
run evidence.research-claims: updating OPEN_QUESTIONS.md, CLAIMS.md, and
DECISIONS.md.research-synthesis: batch/branch review after several experiments.research-references: paper cards and outside evidence.research-director: rank next experiments, decide what not to run, and
select the next scorecard.plan-research-goal: drafting a /goal prompt for autonomous research.aha-moment: personal "now it clicked" notes in AHA/.Keep this skill small. Detailed procedures live in the sub-skills.
RESEARCH.md (its program router names the active direction)programs/index.yml, then the program card the task belongs toOPEN_QUESTIONS.mdCLAIMS.mdDECISIONS.mdexperiments/index.ymlselected scorecard in factory/scorecards/Q question -> H hypothesis -> E experiment -> R run -> C claim -> D decision
Programs (RP###) group this graph by research direction. Resolve which
program the task serves before experiment work; if none is named, work the
highest-priority active program in the RESEARCH.md router.
Each experiment should make one of these moves:
If none of those happened, record it as a scratch run, not a serious experiment.
RESEARCH.md short.aha-moment and capture it in
AHA/.tools/validate_research_workspace.py after structural changes.After about five serious experiments in a branch, use research-synthesis.
The branch should get an explicit continue/stop/pivot decision.
Before starting a new autonomous /goal loop, use research-director if the
next experiment is not already a direct consequence of the latest decision.